Log-Demons with Driving Force for Large Deformation Image Registration

Size: px
Start display at page:

Download "Log-Demons with Driving Force for Large Deformation Image Registration"

Transcription

1 Log-Demons with Driving Force for Large Deformation Image Registration Le Zhang and Ying Wen* Shanghai Key Laboratory of Multidimensional Information Processing, Department of Computer Science and Technology, East China Normal University, Shanghai, China, * Abstract Capturing large diffeomorphic deformations is d- ifficult for many non-rigid registration methods. In this paper, we propose Log-Demons with driving force for large deformation image registration. The driving force obtained by boundary points correspondence exerts influence on continuous optimization of Log-Demons to improve the motion direction of points. We utilize MROGH descriptor matching to obtain points correspondence as driving force, then the driving force is added to the optimization of Log-Demons. We integrate the driving force in an exponentially decreasing form with velocity field of Log-Demons to drive the points moving globally and to speed up the convergence. Experiments performed on synthetic images, real scene images and brain images demonstrate that the proposed method can not only capture large deformations but also preserve details and register images at a higher accuracy. I. INTRODUCTION Non-rigid image registration technique is a significant field of image processing research, which is widely used in military, remote sensing and medicine. Among them, large deformation image registration is a big challenge. Up to now, much work has been done and a large family of algorithms have been put forward[1]-[4]. A predominant way for non-rigid registration is based on Demons algorithm as introduced in the seminal work by Thirion, where he considered image registration as a diffusion process. Thirion considered one image as a deformable model grid, diffusing through the boundaries of the other image which are considered as semi-permeable membranes[2]. The original Demons algorithm is based on gradient information, when the gradient information is low or missing, Demons may lead to wrong registration results. So several kinds of improved Demons algorithms have been proposed[3][4][9]. Vercauteren[16] demonstrated that the above different Demons variants are related to the use of different optimizers. And to improve the search direction of Gauss-Newton-like method, Vercauteren introduced Efficient Second-order Minimization(ESM)[12] into Demons, showing that symmetric forces can speed up convergence rate. However, a widely acknowledged issue with Demons algorithm is that the estimated transformation is not diffeomorphic and may destroy the topological structure which is vital in some practical applications such as computational anatomy. An approach to solve this issue is to compose deformations as a sequence of diffeomorphic free-form deformations modelled by B-splines[6], but the spline transformations become difficult to obtain diffeomorphisms under very large deformation conditions. The large deformation diffeomorphic metric mapping(lddmm)[7][8] was proposed by Beg et al., which seeks a time-dependent velocity field representation of a diffeomorphism and can achieve a high registration accuracy. Since gradient optimization is over the entire space of diffeomorphisms, LDDMM is time consumption and high memory intensive. LDDMM s limitation can be tackled using stationary velocity fields, so Log-Demons[14][15] was proposed by Vercauteren, which is an efficient non-parametric diffeomorphic image registration framework based on Thirion s Demons algorithm. Log-Demons algorithm is appealing as it not only ensures diffeomorphic mappings but also its computation is in linear complexity. Log-Demons achieves diffeomorphism by performing optimization procedure on a Lie group. But for some algorithms based on Log-Demons[5][31][30], it is still difficult to acquire very large and complex deformations. Conventional methods to deal with large deformations are to adopt coarse-to-fine warping framework, but it may still lead gradient-based methods to a local optimal solution, even in coarse resolutions[29]. To solve this problem, many methods based on regional descriptor matching are proposed. Brox[10] and Liu[11] proposed to integrate descriptors such as HOG[22] and dense SIFT[19] into the variational optical flow to preserve details for large displacement. But due to the mismatching of corresponding points, many motion details are still missed and the transformations are not diffeomorphic. Cifor[26] proposed hybrid feature-based Log-Demons registration which combined block-matching scheme with Log-Demons for tumour tracking. The blocks matching affects the motion direction of points in blocks, but it is hard to process images with complex scenario. Lombaert[28][29] also introduced spectral representation of shapes into Log-Demons, using the properties of spectral global matching to capture very large deformations between images, but the computation of spectral coordinates is time-consuming. The above methods, either do not preserve details, or are not diffeomorphic, or time-consuming, so it is not satisfactory for these methods to deal with very large deformation image registration. In order to overcome above mentioned problems, we propose Log-Demons with driving force for large deformation image registration. We first extract images boundaries using Canny algorithm as driving points, and then employ MROGH descriptor to describe each driving point. MROGH descriptor[21] is intrinsic rotation invariant and performs well in points matching and object recognition etc. Hence, the /16/$31.00 c 2016 IEEE 3052

2 characteristics of boundary can be well described by the MROGH descriptor. By descriptor matching, we can obtain corresponding relations which are defined as driving force in this paper. After that, we cast driving force into Log- Demons. Rather than directly replace the displacement vectors with matching results, we add driving force to the velocity field in each iteration to promote the registration. The main contribution of our work is that we employ boundary points correspondence as driving force to affect the update of Log- Demons in a global scope. And we integrate the driving force in an exponentially decreasing form with velocity field of Log- Demons to drive the movements of points and to speed up the convergence. Extensive experiments performed on synthetic images, real scene images and brain images show that our method has an excellent performance, especially for large deformation registration. The rest of the paper is organized as follows: Section II gives a detailed introduction of our method. The experimental results are presented in Section III. Finally, we draw a conclusion in Section IV. II. METHODOLOGY Before we present our method, we give a brief introduction of Log-Demons, since our method derives from Log-Demons. A. Log-Demons Given a fixed image F and a moving image M, Log- Demons estimates a dense diffeomorphic spatial transformation s that aligns M to F best: E(s, c) = 1 λ 2 Sim(F, M c)+ 1 i λ 2 dist(s, c) x λ 2 Reg(s) (1) T where λ i stands for the noise, λ T controls the degree of regularization and λ x stands for a spatial uncertainty on the correspondences. The variable c is exact spatial transformation that allows error for s[3], leading Demons algorithm to a well-posed problem. The similarity term Sim measures the resemblance between fixed image and warped image, and dist describes the distance between two transformations and ensures that s is close to c, and Reg term is used as a priori knowledge and keeps spatial transformation smooth. The transformation s is parameterized by the stationary velocity field v through exponential map s = exp(v). The velocities v is in the space of log-domain. As v is stationary, exp(v) is efficiently computed using the scaling-and-squaring algorithm[14]. The energy of the Log-Demons can be written in log domain: E(u) = 1 λ 2 isim(f, M s exp(u)) + 1 λ 2 dist(s, s exp(u)) x λ 2 Reg(s) T The optimization of energy equation can be decoupled into two steps. The first step is optimizing 1 Sim(F, M λ 2 i s exp(u)) + 1 λ dist(s, s exp(u)) 2 to obtain update field u 2 x with s being given. The update field u is calculated in the Lie algebra and then is mapped in the space of diffeomorphisms via exponential mapping. Thus transformation s is updated in (2) Fig. 1. The sketch of driving force of two corresponding boundary points. (a) Fixed image. (b) Moving image. (c) The correspondence of p and p. (d) The registration result of Log-Demons. the form of s = s exp(u). It should be noted that different optimization strategies lead to different expressions of Demons force[14]. By adapting Gauss-Newton-like approaches, we can get update field in each iteration as: F (x) M s(x) u(x) = J(x) (3) J(x) 2 + λ2 i (x) λ 2 x where J(x) is the gradient, defined as J(x) = M s(x). The second step is usually smooth the update u, using Gaussian kernels K fluid and K diffusion with standard deviations λ fluid and λ diffusion respectively. B. The Proposed Log-Demons with Driving Force 1) Motivation: Capturing very large deformations is a difficult task for many non-rigid image registration since the gradient computation may fall into local optimum. A popular solution is to establish accurate pointwise correspondence by feature matching technique[29] to push points moving in global scope forcibly. Log-Demons is unable to deal with images with very large deformations such as severe image distortion, whose boundaries are hard to match precisely. Considering that the image boundaries are rich in details and critical to diffusing models of Demons, we establish the correspondence between boundary points of two images and take the correspondence as driving force. In our method, We utilize Canny detector to extract boundary points as driving points. The correspondence between driving points of two images can be viewed as the driving force, which is integrated into the Demons force to push the points moving together. To the end, integration of Log-Demons and driving force can lead to a more precise registration. As shown in Fig.1, Fig.1(a) and Fig.1(b) are a fixed image and a moving image, respectively. Fig.1(c) shows the sketch of driving force of two corresponding boundary points. If the registration method can make the point p move to the position of the point p, the result is satisfactory. However, Log-Demons is unable to obtain the exact result shown in Fig.1(d), so we hope the driving force can be added into Demons force to guide the point rightly moving as indicated by the red arrows. In our method, there are two techniques need to be considered: how to integrate the driving force into Log-Demons algorithm and how to get the driving force. We will describe the two techniques in detail in the following sections. 2) Function of Log-Demons with Driving Force: Given a moving image M and a fixed image F, we can obtain the driving force u c of driving points. The u c can be calculated independent of Log-Demons and then integrated into the 2016 International Joint Conference on Neural Networks (IJCNN) 3053

3 continuous optimization of Log-Demons as the driving force. The objective function of our method is as follows: E(u) = Sim(F, M s exp(u) exp( 2 λ i λ uc )) k + 1 λ 2 dist(s, s exp(u) exp( 1 2 x λ uc )) (4) k λ 2 Reg(s) T It is noted that driving force u c is added to the Demons force. u c can be viewed as a constant variable. λ k controls the influence of driving force on update field u and is set as λ k =2 (k 1)/2, where k is the iteration number in Log- Demons. This leads to a high influence at the beginning and decreases rapidly along with iteration, thus the ill effects of inaccurate descriptor matching can be avoided especially when warped image is close to fixed image. Assuming that descriptor matching in global scope has high accuracy, the driving force u c enables global optimization of Log-Demons and pushes points moving towards right direction rapidly and accurately. According to the Baker-Campbell-Hausdorff(BCH) formula[15], the update should be exp(u(u, u c )) exp(u) exp(( 1 λ 2 k u c ), in order to compute simply, we replace it as U(u, u c ) u + 1 λ 2 k u c. Through first order expansion of intensity difference on the first step of Log- Demons[14]: F (x) M s exp(u)(x) F (x) M s(x)+j(x).u(x) (5) and approximation of distance between two diffeomorphisms: dist(s, s exp(u)) u (6) The energy function of first step of Log-Demons can be rewritten as: Algorithm 1 Log-Demons with Driving Force for Large Deformation Image Registration Input: Fixed image F, moving image M and initial velocity field v Output: Transformation s = exp(v) from M to F 1: Use Canny detector on F and M to obtain two boundary point sets as driving points 2: Compute MROGH descriptor M(x p ) for each driving point 3: Obtain correspondence u c between driving points through descriptor matching using Euclidean distance 4: repeat 5: Add u c to the update u mapping M exp(v) to F using Eq.8 6: Smooth the update u K fluid u for fluid-like regularization 7: Update velocity field:v log(exp(v) exp(u)) 8: Smooth velocity field v K diffusion v for diffusionlike regularization 9: until convergence (MROGH)[21] due to its excellent performance in finding corresponding points which have large orientation change. MROGH is intrinsic rotation invariant and do not need to assign a reference orientation for each driving point which may be a major error source for most of the existing methods[21], such as SIFT[19], HOG[22] and DAISY[20], etc. E(u) 1 [ ] F (x) M s(x) 2 Ω x + 0 x Ω x [ J(x) λ i(x) λ x I ] U(u, u c ) (7) where Ω x is the overlap between F and M s. By minimizing the equation Eq.7, at each pixel x, we can get 2 F (x) M s(x) u(x) = J(x) 1 J(x) 2 + λ2 i (x) λ 2 u c (8) λ 2 k x thus, u(x) is our update rule in each iteration. The proposed Log-Demons with driving force for large deformation image registration is summarized in Algorithm 1. 3) Driving Force: In this section, we present the method how to obtain the driving force. We take boundary points obtained by Canny detector as driving points. The information of driving points can be described by the feature descriptor. And then, the points correspondence can be obtained by descriptor matching. In our method, the driving force u c is defined as the correspondence of boundary points. The driving force can adjust points direction of gradient descent in the process of continuously calculating update field u. When describing a point information and performing descriptors matching in large deformation images, we select the Multisupport Region Order Based Gradient Histogram Fig. 2. Description of MROGH descriptor. (a) Construction of local coordinate system. (b) Multisupport Regions. For each driving point, we use a circular region as the support region with n points denoted by R = {x 1,x 2,..., x n }, I(x i ) is the intensity of sample point x i. Intensities of sample points are sorted in nondescending order and then these points are approximate equally divided into g partitions. Then we construct a rotation invariant coordinate system for each sample point when computing descriptors. As shown in Fig.2(a), suppose x p is a driving point and x i is one of the sample points in its support region. Then, we construct local coordinate system by setting x p x i as the positive y-axis and the corresponding vertical line as x-axis for the sample point x i. Local features are calculated in this local coordinate system to obtain rotation invariant. For each sample point x i, we compute gradient in its local International Joint Conference on Neural Networks (IJCNN)

4 Fig. 3. The synthetic images registration results of different methods. From top to bottom, the names and size of images are Heart( ), Lena(85 85), Hand(95 140), Tennis( ), Shoes( ) and Marble( ). And from left to right, images of each row are (a) fixed image, (b) moving image, registration results of (c) Log-Demons, (d) LDDMM, (e) Spectral Log-Demons and (f) Our method respectively. coordinate system: D x (x i )=I(x 1 i ) I(x 5 i ) (9) D y (x i )=I(x 3 i ) I(x 7 i ) (10) where x j i,j =1, 3, 5, 7, are x i s neighboring points along the x-axis and y-axis in the local x-coordinate system. The gradient magnitude m(x i ) and orientation θ(x i ) can be computed as m(x i )= D x (x i ) 2 + D y (x i ) 2 (11) θ(x i )=tan 1 (D y (x i )/D x (x i )) (12) Then the gradient is transformed to a d-dimensional vector, denoted by F G (x i )=(f1 G,f2 G,..., fd G), fj G = m(x i ) (2π/d α(θ(x i), dir j )) (13) 2π/d where dir j =(2π/d) (i 1),i=1, 2,...d, and α(θ(x i ), dir j ) is the angle between θ(x i ) and dir j. In order to reduce the number of incorrect matches, we choose support regions as the N nested regions centered at the driving point with an equal increment of radius as Fig.2(b) shows. In each support region, the local features of all the sample points are then pooled by their intensity orders to form a vector, then we accumulate vectors of different partitions of each support region to represent this support region, defined as D(R) = (F (R 1 ),F(R 2 ),..., F (R g )). F (R i ) is the accumulated vector of partition R i, i.e., F (R i )= F G (x), (14) x R i Finally, all support regions are pooled together to form final descriptor: M(x p )={D 1 D 2...D N }. We use the default parameter as d =8, g =6, N =4, thus the dimensionality of MROGH descriptor is International Joint Conference on Neural Networks (IJCNN) 3055

5 After constructing descriptor for every driving point, for point x i in moving image, we try to find its corresponding point x j in fixed image by MROGH descriptor matching. The point in the fixed image has a corresponding point in the moving image followed by that x i is the optimal correspondence for x j and x j is also the optimal correspondence to the other. If x i and x j are mutually optimal corresponding points, the displacement vector of point x i is obtained by u c i = x j x i, otherwise u c i =0. III. EXPERIMENTS In this section, we perform experiments on three types of images, i.e., synthetic images, real scene images and brain images to investigate the performance of our proposed method. We compare the proposed method with some popular registration algorithms, such as Log-Demons, Spectral Log-Demons and LDDMM. In the following, we use the default parameters of LDDMM, and all experiments are carried on the following empirical parameters: λ i =1, λ fluid =1, λ diffusion =1. We utilize Mean Square Error(MSE) to evaluate the intensity difference between the fixed image and the warped image. A. Experiments on Synthetic Images In this experiment, we focus on image registration with very large deformations. The moving images are all synthetic images. The max step λ x is set as λ x =2. After performing 100 iterations, the methods have been reached convergence, so we compare them within the same level of resolution. We select six fixed images and warp them with randomly large deformations, and the final registration results are shown in Fig.3. The first image of each row is a fixed image and the second is a deformed moving image, and the followings are registration results of Log-Demons, LDDMM, Spectral Log- Demons, and ours respectively. From Fig.3, it can be seen that in example of Heart, the moving image has some severely distorted parts marked by red arrows. Log-Demons, LDDMM and Spectral Log-Demons register unsuccessfully in these regions, while our method outperforms the other three methods and pushes all boundaries aligned. For the images of Hand, ring finger in the moving image and middle finger in the fixed image are overlapped. LDDMM and Spectral Log-Demons are not capable of moving fingers back to original position. Log-Demons wrongly aligns ring finger to middle finger, resulting in overlap and discontinuous of fingers, since the optimization may be limited in local scope. Due to the correspondence of boundary points, driving force of our method can push the finger back to the right position. The final MSE result of LDDMM is , Log- Demons is , Spectral Log-Demons is and ours is The intensity difference between fixed and warped images of our method is reduced by 59.66% compared with Log-Demons, 57.93% compared with Spectral Log-Demons and 86% compared with LDDMM. As for Lena, Log-Demons, Spectral Log-Demons and LDDMM can not align face or hair, while our method registers all correctly. For Tennis and Marble, twisted lines are hard task for registration methods. For example, the railings in Tennis are moved to wrong places by the other three methods since the lack of external force pulls them aligned with corresponding distant lines. While our method restores the lines successfully, demonstrating that Fig. 5. The colormaps of intensity difference for Hand. Blue means no difference and red means that intensity difference is biggest. From top to bottom, colormaps are obtained by Log-Demons, LDDMM, Spectral Log- Demons and ours respectively when iteration numbers are 1, 5, 10, 20, 100. driving force obtained by boundary points correspondence can restrain the movement of points and is crucial to keep complete object outline and details in the process of registration. In order to show more details in the process of registration, we take Hand as example to show intensity differences in refine iterations. For the sake of visual observation, we transform the intensity differences of Hand into colormap. Fig.5 shows the colormap of warped images when iteration numbers are 1, 5, 10, 20, 100. In the first 10 iterations, the fingers are nearly correctly restored by our method while others fail. It is noted that the driving force should have a great impact on Log-Demons at the beginning, and declines with the decrease of the intensity difference. The MSE curves of registration of Fig.3 are shown in Fig.4, in which the blue, black, green and red curves stand for Spectral Log-Demons, Log-Demons, LDDMM and our method respectively. For six experiments, our MSE curve drops fastest to approach convergence and is nearly always bellow the other methods, indicating that our method registers effectively and accurately. B. Experiments on Real Scene Images In this experiment, we investigate the performance of our proposed method on real scene images. We randomly choose 30 images from MIT[35] database and 20 images from Brox datasets[10] to evaluate. This experiment is conducted in four levels scheme. Due to the limitation of paper length, we present some sample images and registration results in Fig.6. The first image of each row is a fixed image and the second image is a moving image, and the followings are registration results of Log-Demons, LDDMM, Spectral Log-Demons, and ours respectively. As shown in Fig.6, all final registration results are very close in vision, but MSE of ours is superior to that of Log- Demons and Spectral Log-Demons, and LDDMM performs unstably. Table I presents the mean and standard deviation of MSE. It can be seen that the performance of our method on International Joint Conference on Neural Networks (IJCNN)

6 Fig. 4. MSE curves of different registration methods, the blue, black, green and red curves stand for Spectral Log-Demons, Log-Demons, LDDMM and our method, respectively. real scene images still has advantages, while LDDMM has large fluctuations. We can conclude that the performance of our method is superior to Log-Demons based methods on real scene images, and the driving force integrated into Log- Demons can lead to a more precise result. Mean MSE TABLE I. EXPERIMENTS ON REAL SCENE IMAGES Log-Demons LDDMM Spectral Log-Demons Ours Hand 33.95(±16.80) 21.65(±6.08) 36.20(±20.84) 32.27(±14.54) Table (±8.66) 84.35(±7.82) (±6.88) (±6.08) Toys (±41.97) (±80.11) (±40.02) (±38.28) Tennis (±9.22) (±14.65) (±8.54) (±10.61) Shoes (±12.66) (±85.53) (±11.17) (±15.49) C. Experiments on MR Images Demons algorithm is widely used in brain image registration, so we experimentally evaluate the performance of proposed scheme on MR brain images. The brains present complex structures across individuals, so we focus on the warping accuracy of the tissue structure in this experiments. We apply the proposed method to MR brain images registration and 35 T1 images[33] are randomly selected for evaluation. These brains vary significantly in morphology. We also conduct the experiment in four levels scheme and the maximal step is adjusted to 1.5. Fig.7 shows the registration results, the first image of each row is a fixed image and second is a moving image, and the followings are final results of Log-Demons, LDDMM, Spectral Log-Demons and ours respectively. The registration results of all methods have similar observations. In order to evaluate the registration accuracy, the brain image is segmented into three tissues: white matter(wm), gray matter(gm), and cerebrospinal fluid(csf) by the method of FSL[36]. Then, we measured MSE and Dice coefficient[34] to estimate the accuracy, and the Dice coefficient is defined as: ρ =2 O 1 O 2 /( O 1 + O 2 ) (15) where O1 and O2 denote region of interest (ROI) in the fixed and warped images. The MSE and dice coefficient of the four methods are listed in Table II. Our method has a slightly advantage over Log-Demons and Spectral Log- Demons in GM and CSF, while LDDMM performs the worst. The experimental results show that our method is comparable to other state of the art registration methods. IV. CONCLUSION In this paper, we propose Log-Demons with driving force for large deformation image registration. The correspondence of boundary points obtained by MROGH descriptor matching is defined as the driving force which is integrated into Log- Demons to improve the update way of Log-Demons. Thus the proposed method can not only capture large deformations but also preserve details and register images at a higher accuracy. Experiments on synthetic images, real scene images and brain images demonstrate that our method has a good performance for image registration. In particular, our method is superior to other methods for large deformation image registration. ACKNOWLEDGMENT The authors would like to thank Herve Lombaert, Laurent Risser and Bin Fan for sharing their code. This work was supported by The National Natural Science Foundation of China (No ), Shanghai Collaborative Innovation Center of Trustworthy Software for Internet of Things (ZF1213), Open Project of the Key Laboratory of Embedded System and Service Computing Ministry of Education (Tongji University) 2016 International Joint Conference on Neural Networks (IJCNN) 3057

7 Fig. 6. Experiments on real scene images. The first image of each row is a fixed image, and the second is a moving image. The followings are registration results of Log-Demons, LDDMM, Spectral Log-Demons, and Ours respectively. MSE is below images. Fig. 7. Experiments on MR brain images. The first image of each row is a fixed image, and the second is a moving image. The followings are registration results of Log-Demons, LDDMM, Spectral Log-Demons, and Ours respectively International Joint Conference on Neural Networks (IJCNN)

8 TABLE II. EXPERIMENTS ON BRAIN IMAGE REGISTRATION Dice Log-Demons LDDMM Spectral Log-Demons Ours coefficient WM (±0.0656) (±0.0778) (±0.0662) (±0.0655) GM (±0.0153) (±0.0170) (±0.0159) (±0.0155) CSF (±0.0242) (±0.0286) (±0.0235) (±0.0235) MSE (± ) (± ) (± ) (± ) and the Science and Technology Commission of Shanghai Municipality under research grant no.14dz REFERENCES [1] G. E. Christensen, P. Yin, M. W. Vannier, et al, Large-deformation image registration using fluid landmarks, Image Analysis and Interpretation, Proceedings. 4th IEEE Southwest Symposium. pp , [2] J. P. Thirion, Image matching as a diffusion process: an analogy with Maxwell s demons, Medical Image Analysis, vol. 2, pp , [3] P. Cachier, E. Bardinet, D. Dormont, et al, Iconic Feature Based Nonrigid Registration: The PASHA Algorithm, Computer Vision and Image Understanding, vol. 89, pp , [4] X. Pennec, P. Cachier, N. Ayache, Understanding the Demons Algorithm: 3D Non-rigid Registration by Gradient Descent, Medical Image Computing and Computer-Assisted Intervention-MICCAI99, pp , [5] T. Mansi, X. Pennec, M. Sermesant, et al, ilogdemons: A Demons- Based Registration Algorithm for Tracking Incompressible Elastic Biological Tissues, International Journal of Computer Vision, vol. 92, pp , [6] D. Rueckert, P. Aljabar, R. A. Heckemann, et al, Diffeomorphic registration using B-splines, 9th International Conference on Medical Image Computing and Computer-Assisted Intervention: MICCAI, pp , [7] M. F. Beg, M. I. Miller, A. Trouve, and L. Younes, Computing large deformation metric mappings via geodesic flows of diffeomorphisms, International journal of computer vision, pp , [8] X. Yang, Y. Li, D. Reutens0, et al, Diffeomorphic Metric Landmark Mapping Using Stationary Velocity Field Parameterization, International Journal of Computer Vision, pp. 1-18, [9] H. Wang, L. Dong, J. O Daniel, et al, Validation of an accelerated demons algorithm for deformable image registration in radiation therapy, Physics in Medicine and Biology, vol. 50, pp , [10] T. Brox, J. Malik, Large Displacement Optical Flow: Descriptor Matching in Variational Motion Estimation, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 33, pp , [11] C. Liu, J. Yuen, A. Torralba, J. Sivic, and W. T. Freeman, SIFT Flow: Dense Correspondence across Different Scenes, Proc. European Conf. Computer Vision, pp , [12] S. Benhimane, E. Malis, Real-time image-based tracking of planes using efficient second-order minimization, Proceedings of the International Conference on Intelligent Robots and Systems, vol. 1, pp , [13] J. Ashburner, A fast diffeomorphic image registration algorithm, Neuroimage, vol. 38, pp , [14] T. Vercauteren, P. Xavier, P. Aymeric, et al, Diffeomorphic demons: efficient non-parametric image registration, Neuroimage, vol. 45, pp , [15] T. Vercauteren, X. Pennec, A. Perchant, N. Ayache, Symmetric Log- Domain diffeomorphic registration: A Demons-Based approach, MIC- CAI, pp , [16] T. Vercauteren, X. Pennec, E. Malis, A. Perchant, N. Ayache, Insight into efficient image registration techniques and the demons algorithm, In Information Processing in Medical Imaging, pp , [17] D. Florence, T. Vercauteren, An ITK Implementation of the Symmetric Log-Domain Diffeomorphic Demons Algorithm, Insight Journal, [18] B. T. Yeo, M. R. Sabuncu, T. Vercauteren, et al, Spherical demons: fast diffeomorphic landmark-free surface registration, IEEE Transactions on Medical Imaging, vol. 29, pp , [19] D. G. Lowe, Distinctive Image Features from Scale-Invariant Keypoints, International Journal of Computer Vision, vol. 60, pp, , [20] E. Tola, V. Lepetit, P. Fua, A fast local descriptor for dense matching, IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 1-8, [21] B. Fan, F. Wu, Z. Hu, Rotationally Invariant Descriptors Using Intensity Order Pooling, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 34, pp , [22] N. Dalal, B. Triggs, Histograms of Oriented Gradients for Human Detection, Computer Vision and Pattern Recognition, vol. 1, pp , [23] J. Ma, W. Qiu, J. Zhao, et al, Robust Estimation of Transformation for Non-Rigid Registration, IEEE Transactions on Signal Processing, vol. 63, pp , [24] A. Myronenko, X. Song, Point Set Registration: Coherent Point Drift, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 32, pp , [25] H. Chui, A. Rangarajan, A new point matching algorithm for nonrigid registration, Computer Vision and Image Understanding, vol. 89, pp , [26] A. Cifor, L. Risser L, D. Chung, et al, Hybrid feature-based Log- Demons registration for tumour tracking in 2-D liver ultrasound images, IEEE International Symposium on Biomedical Imaging, pp , [27] M. FreimanS. D. VossS. K. Warfield, Demons registration with local affine adaptive regularization: application to registration of abdominal structures, Proceedings of the 8th IEEE International Symposium on Biomedical Imaging, pp , [28] H. Lombaert, L. Grady, X. Pennec, et al, Spectral Demons-Image Registration via Global Spectral Correspondence, Computer Vision- ECCV 2012, pp , [29] H. Lombaert, L. Grady, X. Pennec, et al, Spectral Log-Demons: Diffeomorphic Image Registration with Very Large Deformations, International Journal of Computer Vision, vol. 107, pp , [30] C. Seiler, X. Pennec, M. Reyes, Geometry-Aware Multiscale Image Registration Via OBBTree-Based Polyaffine Log-Demons, Lecture Notes in Computer Science, pp , [31] V. Siless, J. Glauns, P. Guevara, et al, Joint T1 and Brain Fiber Log-Demons Registration Using Currents to Model Geometry, Medical Image Computing and Computer-Assisted Intervention-MICCAI 2012, pp , [32] L. Risser, F. Vialard, R. Wolz, et al, Simultaneous multi-scale registration using large deformation diffeomorphic metric mapping, IEEE Transactions on Medical Imaging, vol. 30, pp , [33] [34] H. H. Chang, A. H. Zhuang, D. J. Valentino, and W. C. Chu, Performance measure characterization for evaluating neuroimage segmentation algorithms, Neuroimage, vol. 47, pp , [35] [36] M. Jenkinson, C. F. Beckmann, T. E. J. Behrens, et al, FSL, Neuroimage, vol. 62, pp , International Joint Conference on Neural Networks (IJCNN) 3059

Visual Tracking (1) Tracking of Feature Points and Planar Rigid Objects

Visual Tracking (1) Tracking of Feature Points and Planar Rigid Objects Intelligent Control Systems Visual Tracking (1) Tracking of Feature Points and Planar Rigid Objects Shingo Kagami Graduate School of Information Sciences, Tohoku University swk(at)ic.is.tohoku.ac.jp http://www.ic.is.tohoku.ac.jp/ja/swk/

More information

A Novel Extreme Point Selection Algorithm in SIFT

A Novel Extreme Point Selection Algorithm in SIFT A Novel Extreme Point Selection Algorithm in SIFT Ding Zuchun School of Electronic and Communication, South China University of Technolog Guangzhou, China zucding@gmail.com Abstract. This paper proposes

More information

ADAPTIVE GRAPH CUTS WITH TISSUE PRIORS FOR BRAIN MRI SEGMENTATION

ADAPTIVE GRAPH CUTS WITH TISSUE PRIORS FOR BRAIN MRI SEGMENTATION ADAPTIVE GRAPH CUTS WITH TISSUE PRIORS FOR BRAIN MRI SEGMENTATION Abstract: MIP Project Report Spring 2013 Gaurav Mittal 201232644 This is a detailed report about the course project, which was to implement

More information

Chapter 3 Image Registration. Chapter 3 Image Registration

Chapter 3 Image Registration. Chapter 3 Image Registration Chapter 3 Image Registration Distributed Algorithms for Introduction (1) Definition: Image Registration Input: 2 images of the same scene but taken from different perspectives Goal: Identify transformation

More information

Adaptive Zoom Distance Measuring System of Camera Based on the Ranging of Binocular Vision

Adaptive Zoom Distance Measuring System of Camera Based on the Ranging of Binocular Vision Adaptive Zoom Distance Measuring System of Camera Based on the Ranging of Binocular Vision Zhiyan Zhang 1, Wei Qian 1, Lei Pan 1 & Yanjun Li 1 1 University of Shanghai for Science and Technology, China

More information

Non-rigid Image Registration using Electric Current Flow

Non-rigid Image Registration using Electric Current Flow Non-rigid Image Registration using Electric Current Flow Shu Liao, Max W. K. Law and Albert C. S. Chung Lo Kwee-Seong Medical Image Analysis Laboratory, Department of Computer Science and Engineering,

More information

Fast Image Registration via Joint Gradient Maximization: Application to Multi-Modal Data

Fast Image Registration via Joint Gradient Maximization: Application to Multi-Modal Data MITSUBISHI ELECTRIC RESEARCH LABORATORIES http://www.merl.com Fast Image Registration via Joint Gradient Maximization: Application to Multi-Modal Data Xue Mei, Fatih Porikli TR-19 September Abstract We

More information

Learning-based Neuroimage Registration

Learning-based Neuroimage Registration Learning-based Neuroimage Registration Leonid Teverovskiy and Yanxi Liu 1 October 2004 CMU-CALD-04-108, CMU-RI-TR-04-59 School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 Abstract

More information

Shape-based Diffeomorphic Registration on Hippocampal Surfaces Using Beltrami Holomorphic Flow

Shape-based Diffeomorphic Registration on Hippocampal Surfaces Using Beltrami Holomorphic Flow Shape-based Diffeomorphic Registration on Hippocampal Surfaces Using Beltrami Holomorphic Flow Abstract. Finding meaningful 1-1 correspondences between hippocampal (HP) surfaces is an important but difficult

More information

A Novel Image Super-resolution Reconstruction Algorithm based on Modified Sparse Representation

A Novel Image Super-resolution Reconstruction Algorithm based on Modified Sparse Representation , pp.162-167 http://dx.doi.org/10.14257/astl.2016.138.33 A Novel Image Super-resolution Reconstruction Algorithm based on Modified Sparse Representation Liqiang Hu, Chaofeng He Shijiazhuang Tiedao University,

More information

Video annotation based on adaptive annular spatial partition scheme

Video annotation based on adaptive annular spatial partition scheme Video annotation based on adaptive annular spatial partition scheme Guiguang Ding a), Lu Zhang, and Xiaoxu Li Key Laboratory for Information System Security, Ministry of Education, Tsinghua National Laboratory

More information

Subpixel Corner Detection Using Spatial Moment 1)

Subpixel Corner Detection Using Spatial Moment 1) Vol.31, No.5 ACTA AUTOMATICA SINICA September, 25 Subpixel Corner Detection Using Spatial Moment 1) WANG She-Yang SONG Shen-Min QIANG Wen-Yi CHEN Xing-Lin (Department of Control Engineering, Harbin Institute

More information

Image Segmentation and Registration

Image Segmentation and Registration Image Segmentation and Registration Dr. Christine Tanner (tanner@vision.ee.ethz.ch) Computer Vision Laboratory, ETH Zürich Dr. Verena Kaynig, Machine Learning Laboratory, ETH Zürich Outline Segmentation

More information

Motion Estimation. There are three main types (or applications) of motion estimation:

Motion Estimation. There are three main types (or applications) of motion estimation: Members: D91922016 朱威達 R93922010 林聖凱 R93922044 謝俊瑋 Motion Estimation There are three main types (or applications) of motion estimation: Parametric motion (image alignment) The main idea of parametric motion

More information

Thin Plate Spline Feature Point Matching for Organ Surfaces in Minimally Invasive Surgery Imaging

Thin Plate Spline Feature Point Matching for Organ Surfaces in Minimally Invasive Surgery Imaging Thin Plate Spline Feature Point Matching for Organ Surfaces in Minimally Invasive Surgery Imaging Bingxiong Lin, Yu Sun and Xiaoning Qian University of South Florida, Tampa, FL., U.S.A. ABSTRACT Robust

More information

FAST REGISTRATION OF TERRESTRIAL LIDAR POINT CLOUD AND SEQUENCE IMAGES

FAST REGISTRATION OF TERRESTRIAL LIDAR POINT CLOUD AND SEQUENCE IMAGES FAST REGISTRATION OF TERRESTRIAL LIDAR POINT CLOUD AND SEQUENCE IMAGES Jie Shao a, Wuming Zhang a, Yaqiao Zhu b, Aojie Shen a a State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing

More information

Structured Light II. Thanks to Ronen Gvili, Szymon Rusinkiewicz and Maks Ovsjanikov

Structured Light II. Thanks to Ronen Gvili, Szymon Rusinkiewicz and Maks Ovsjanikov Structured Light II Johannes Köhler Johannes.koehler@dfki.de Thanks to Ronen Gvili, Szymon Rusinkiewicz and Maks Ovsjanikov Introduction Previous lecture: Structured Light I Active Scanning Camera/emitter

More information

3D Computer Vision. Structured Light II. Prof. Didier Stricker. Kaiserlautern University.

3D Computer Vision. Structured Light II. Prof. Didier Stricker. Kaiserlautern University. 3D Computer Vision Structured Light II Prof. Didier Stricker Kaiserlautern University http://ags.cs.uni-kl.de/ DFKI Deutsches Forschungszentrum für Künstliche Intelligenz http://av.dfki.de 1 Introduction

More information

EE795: Computer Vision and Intelligent Systems

EE795: Computer Vision and Intelligent Systems EE795: Computer Vision and Intelligent Systems Spring 2012 TTh 17:30-18:45 FDH 204 Lecture 14 130307 http://www.ee.unlv.edu/~b1morris/ecg795/ 2 Outline Review Stereo Dense Motion Estimation Translational

More information

Visual Tracking (1) Pixel-intensity-based methods

Visual Tracking (1) Pixel-intensity-based methods Intelligent Control Systems Visual Tracking (1) Pixel-intensity-based methods Shingo Kagami Graduate School of Information Sciences, Tohoku University swk(at)ic.is.tohoku.ac.jp http://www.ic.is.tohoku.ac.jp/ja/swk/

More information

Fully Automatic Multi-organ Segmentation based on Multi-boost Learning and Statistical Shape Model Search

Fully Automatic Multi-organ Segmentation based on Multi-boost Learning and Statistical Shape Model Search Fully Automatic Multi-organ Segmentation based on Multi-boost Learning and Statistical Shape Model Search Baochun He, Cheng Huang, Fucang Jia Shenzhen Institutes of Advanced Technology, Chinese Academy

More information

SUMMARY: DISTINCTIVE IMAGE FEATURES FROM SCALE- INVARIANT KEYPOINTS

SUMMARY: DISTINCTIVE IMAGE FEATURES FROM SCALE- INVARIANT KEYPOINTS SUMMARY: DISTINCTIVE IMAGE FEATURES FROM SCALE- INVARIANT KEYPOINTS Cognitive Robotics Original: David G. Lowe, 004 Summary: Coen van Leeuwen, s1460919 Abstract: This article presents a method to extract

More information

CS 231A Computer Vision (Fall 2012) Problem Set 3

CS 231A Computer Vision (Fall 2012) Problem Set 3 CS 231A Computer Vision (Fall 2012) Problem Set 3 Due: Nov. 13 th, 2012 (2:15pm) 1 Probabilistic Recursion for Tracking (20 points) In this problem you will derive a method for tracking a point of interest

More information

An Algorithm for Medical Image Registration using Local Feature Modal Mapping

An Algorithm for Medical Image Registration using Local Feature Modal Mapping An Algorithm for Medical Image Registration using Local Feature Modal Mapping Cundong Tang, Shangke Quan,Xinfeng Yang * School of Computer and Information Engineering, Nanyang Institute of Technology,

More information

Face Tracking. Synonyms. Definition. Main Body Text. Amit K. Roy-Chowdhury and Yilei Xu. Facial Motion Estimation

Face Tracking. Synonyms. Definition. Main Body Text. Amit K. Roy-Chowdhury and Yilei Xu. Facial Motion Estimation Face Tracking Amit K. Roy-Chowdhury and Yilei Xu Department of Electrical Engineering, University of California, Riverside, CA 92521, USA {amitrc,yxu}@ee.ucr.edu Synonyms Facial Motion Estimation Definition

More information

Methodological progress in image registration for ventilation estimation, segmentation propagation and multi-modal fusion

Methodological progress in image registration for ventilation estimation, segmentation propagation and multi-modal fusion Methodological progress in image registration for ventilation estimation, segmentation propagation and multi-modal fusion Mattias P. Heinrich Julia A. Schnabel, Mark Jenkinson, Sir Michael Brady 2 Clinical

More information

Regional Manifold Learning for Deformable Registration of Brain MR Images

Regional Manifold Learning for Deformable Registration of Brain MR Images Regional Manifold Learning for Deformable Registration of Brain MR Images Dong Hye Ye, Jihun Hamm, Dongjin Kwon, Christos Davatzikos, and Kilian M. Pohl Department of Radiology, University of Pennsylvania,

More information

Feature Detection. Raul Queiroz Feitosa. 3/30/2017 Feature Detection 1

Feature Detection. Raul Queiroz Feitosa. 3/30/2017 Feature Detection 1 Feature Detection Raul Queiroz Feitosa 3/30/2017 Feature Detection 1 Objetive This chapter discusses the correspondence problem and presents approaches to solve it. 3/30/2017 Feature Detection 2 Outline

More information

Peripheral drift illusion

Peripheral drift illusion Peripheral drift illusion Does it work on other animals? Computer Vision Motion and Optical Flow Many slides adapted from J. Hays, S. Seitz, R. Szeliski, M. Pollefeys, K. Grauman and others Video A video

More information

Curvature guided surface registration using level sets

Curvature guided surface registration using level sets Curvature guided surface registration using level sets Marcel Lüthi, Thomas Albrecht, Thomas Vetter Department of Computer Science, University of Basel, Switzerland Abstract. We present a new approach

More information

ELEC Dr Reji Mathew Electrical Engineering UNSW

ELEC Dr Reji Mathew Electrical Engineering UNSW ELEC 4622 Dr Reji Mathew Electrical Engineering UNSW Review of Motion Modelling and Estimation Introduction to Motion Modelling & Estimation Forward Motion Backward Motion Block Motion Estimation Motion

More information

CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION

CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION 6.1 INTRODUCTION Fuzzy logic based computational techniques are becoming increasingly important in the medical image analysis arena. The significant

More information

Segmentation and Tracking of Partial Planar Templates

Segmentation and Tracking of Partial Planar Templates Segmentation and Tracking of Partial Planar Templates Abdelsalam Masoud William Hoff Colorado School of Mines Colorado School of Mines Golden, CO 800 Golden, CO 800 amasoud@mines.edu whoff@mines.edu Abstract

More information

ABSTRACT 1. INTRODUCTION 2. METHODS

ABSTRACT 1. INTRODUCTION 2. METHODS Finding Seeds for Segmentation Using Statistical Fusion Fangxu Xing *a, Andrew J. Asman b, Jerry L. Prince a,c, Bennett A. Landman b,c,d a Department of Electrical and Computer Engineering, Johns Hopkins

More information

Non-Rigid Image Registration III

Non-Rigid Image Registration III Non-Rigid Image Registration III CS6240 Multimedia Analysis Leow Wee Kheng Department of Computer Science School of Computing National University of Singapore Leow Wee Kheng (CS6240) Non-Rigid Image Registration

More information

The Anatomical Equivalence Class Formulation and its Application to Shape-based Computational Neuroanatomy

The Anatomical Equivalence Class Formulation and its Application to Shape-based Computational Neuroanatomy The Anatomical Equivalence Class Formulation and its Application to Shape-based Computational Neuroanatomy Sokratis K. Makrogiannis, PhD From post-doctoral research at SBIA lab, Department of Radiology,

More information

IMAGE RETRIEVAL USING VLAD WITH MULTIPLE FEATURES

IMAGE RETRIEVAL USING VLAD WITH MULTIPLE FEATURES IMAGE RETRIEVAL USING VLAD WITH MULTIPLE FEATURES Pin-Syuan Huang, Jing-Yi Tsai, Yu-Fang Wang, and Chun-Yi Tsai Department of Computer Science and Information Engineering, National Taitung University,

More information

Nonrigid Surface Modelling. and Fast Recovery. Department of Computer Science and Engineering. Committee: Prof. Leo J. Jia and Prof. K. H.

Nonrigid Surface Modelling. and Fast Recovery. Department of Computer Science and Engineering. Committee: Prof. Leo J. Jia and Prof. K. H. Nonrigid Surface Modelling and Fast Recovery Zhu Jianke Supervisor: Prof. Michael R. Lyu Committee: Prof. Leo J. Jia and Prof. K. H. Wong Department of Computer Science and Engineering May 11, 2007 1 2

More information

Feature Tracking and Optical Flow

Feature Tracking and Optical Flow Feature Tracking and Optical Flow Prof. D. Stricker Doz. G. Bleser Many slides adapted from James Hays, Derek Hoeim, Lana Lazebnik, Silvio Saverse, who 1 in turn adapted slides from Steve Seitz, Rick Szeliski,

More information

Traffic Signs Recognition using HP and HOG Descriptors Combined to MLP and SVM Classifiers

Traffic Signs Recognition using HP and HOG Descriptors Combined to MLP and SVM Classifiers Traffic Signs Recognition using HP and HOG Descriptors Combined to MLP and SVM Classifiers A. Salhi, B. Minaoui, M. Fakir, H. Chakib, H. Grimech Faculty of science and Technology Sultan Moulay Slimane

More information

Improving Latent Fingerprint Matching Performance by Orientation Field Estimation using Localized Dictionaries

Improving Latent Fingerprint Matching Performance by Orientation Field Estimation using Localized Dictionaries Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 3, Issue. 11, November 2014,

More information

Object Tracking using HOG and SVM

Object Tracking using HOG and SVM Object Tracking using HOG and SVM Siji Joseph #1, Arun Pradeep #2 Electronics and Communication Engineering Axis College of Engineering and Technology, Ambanoly, Thrissur, India Abstract Object detection

More information

Algorithm research of 3D point cloud registration based on iterative closest point 1

Algorithm research of 3D point cloud registration based on iterative closest point 1 Acta Technica 62, No. 3B/2017, 189 196 c 2017 Institute of Thermomechanics CAS, v.v.i. Algorithm research of 3D point cloud registration based on iterative closest point 1 Qian Gao 2, Yujian Wang 2,3,

More information

Occluded Facial Expression Tracking

Occluded Facial Expression Tracking Occluded Facial Expression Tracking Hugo Mercier 1, Julien Peyras 2, and Patrice Dalle 1 1 Institut de Recherche en Informatique de Toulouse 118, route de Narbonne, F-31062 Toulouse Cedex 9 2 Dipartimento

More information

3D Surface Matching and Registration through Shape Images

3D Surface Matching and Registration through Shape Images 3D Surface Matching and Registration through Shape Images Zhaoqiang Lai and Jing Hua Department of Computer Science, Wayne State University, Detroit, MI 48202, USA Abstract. In this paper, we present a

More information

EECS150 - Digital Design Lecture 14 FIFO 2 and SIFT. Recap and Outline

EECS150 - Digital Design Lecture 14 FIFO 2 and SIFT. Recap and Outline EECS150 - Digital Design Lecture 14 FIFO 2 and SIFT Oct. 15, 2013 Prof. Ronald Fearing Electrical Engineering and Computer Sciences University of California, Berkeley (slides courtesy of Prof. John Wawrzynek)

More information

Visual Tracking (1) Feature Point Tracking and Block Matching

Visual Tracking (1) Feature Point Tracking and Block Matching Intelligent Control Systems Visual Tracking (1) Feature Point Tracking and Block Matching Shingo Kagami Graduate School of Information Sciences, Tohoku University swk(at)ic.is.tohoku.ac.jp http://www.ic.is.tohoku.ac.jp/ja/swk/

More information

Correspondence Detection Using Wavelet-Based Attribute Vectors

Correspondence Detection Using Wavelet-Based Attribute Vectors Correspondence Detection Using Wavelet-Based Attribute Vectors Zhong Xue, Dinggang Shen, and Christos Davatzikos Section of Biomedical Image Analysis, Department of Radiology University of Pennsylvania,

More information

Rotation Invariant Image Registration using Robust Shape Matching

Rotation Invariant Image Registration using Robust Shape Matching International Journal of Electronic and Electrical Engineering. ISSN 0974-2174 Volume 7, Number 2 (2014), pp. 125-132 International Research Publication House http://www.irphouse.com Rotation Invariant

More information

Epithelial rosette detection in microscopic images

Epithelial rosette detection in microscopic images Epithelial rosette detection in microscopic images Kun Liu,3, Sandra Ernst 2,3, Virginie Lecaudey 2,3 and Olaf Ronneberger,3 Department of Computer Science 2 Department of Developmental Biology 3 BIOSS

More information

CS 4495 Computer Vision Motion and Optic Flow

CS 4495 Computer Vision Motion and Optic Flow CS 4495 Computer Vision Aaron Bobick School of Interactive Computing Administrivia PS4 is out, due Sunday Oct 27 th. All relevant lectures posted Details about Problem Set: You may *not* use built in Harris

More information

Deformable Registration Using Scale Space Keypoints

Deformable Registration Using Scale Space Keypoints Deformable Registration Using Scale Space Keypoints Mehdi Moradi a, Purang Abolmaesoumi a,b and Parvin Mousavi a a School of Computing, Queen s University, Kingston, Ontario, Canada K7L 3N6; b Department

More information

EXAM SOLUTIONS. Image Processing and Computer Vision Course 2D1421 Monday, 13 th of March 2006,

EXAM SOLUTIONS. Image Processing and Computer Vision Course 2D1421 Monday, 13 th of March 2006, School of Computer Science and Communication, KTH Danica Kragic EXAM SOLUTIONS Image Processing and Computer Vision Course 2D1421 Monday, 13 th of March 2006, 14.00 19.00 Grade table 0-25 U 26-35 3 36-45

More information

arxiv: v1 [cs.cv] 28 Sep 2018

arxiv: v1 [cs.cv] 28 Sep 2018 Camera Pose Estimation from Sequence of Calibrated Images arxiv:1809.11066v1 [cs.cv] 28 Sep 2018 Jacek Komorowski 1 and Przemyslaw Rokita 2 1 Maria Curie-Sklodowska University, Institute of Computer Science,

More information

Integrated Approaches to Non-Rigid Registration in Medical Images

Integrated Approaches to Non-Rigid Registration in Medical Images Work. on Appl. of Comp. Vision, pg 102-108. 1 Integrated Approaches to Non-Rigid Registration in Medical Images Yongmei Wang and Lawrence H. Staib + Departments of Electrical Engineering and Diagnostic

More information

VALIDATION OF DIR. Raj Varadhan, PhD, DABMP Minneapolis Radiation Oncology

VALIDATION OF DIR. Raj Varadhan, PhD, DABMP Minneapolis Radiation Oncology VALIDATION OF DIR Raj Varadhan, PhD, DABMP Minneapolis Radiation Oncology Overview Basics: Registration Framework, Theory Discuss Validation techniques Using Synthetic CT data & Phantoms What metrics to

More information

MULTIPLE REGION OF INTEREST TRACKING OF NON-RIGID OBJECTS USING DEMON'S ALGORITHM

MULTIPLE REGION OF INTEREST TRACKING OF NON-RIGID OBJECTS USING DEMON'S ALGORITHM MULTIPLE REGION OF INTEREST TRACKING OF NON-RIGID OBJECTS USING DEMON'S ALGORITHM Rohan Pillai 1, Abhishikta Yalavali 2 Saima Mohan 1, and Amol Patil 1 1 igate, Tower 3, Magarpatta City, Pune, India, {rohan.pillai,saima.mohan,amol.vpatil}@igate.com,

More information

An Adaptive Eigenshape Model

An Adaptive Eigenshape Model An Adaptive Eigenshape Model Adam Baumberg and David Hogg School of Computer Studies University of Leeds, Leeds LS2 9JT, U.K. amb@scs.leeds.ac.uk Abstract There has been a great deal of recent interest

More information

Translation Symmetry Detection: A Repetitive Pattern Analysis Approach

Translation Symmetry Detection: A Repetitive Pattern Analysis Approach 2013 IEEE Conference on Computer Vision and Pattern Recognition Workshops Translation Symmetry Detection: A Repetitive Pattern Analysis Approach Yunliang Cai and George Baciu GAMA Lab, Department of Computing

More information

Sea Turtle Identification by Matching Their Scale Patterns

Sea Turtle Identification by Matching Their Scale Patterns Sea Turtle Identification by Matching Their Scale Patterns Technical Report Rajmadhan Ekambaram and Rangachar Kasturi Department of Computer Science and Engineering, University of South Florida Abstract

More information

Comparative Study of ROI Extraction of Palmprint

Comparative Study of ROI Extraction of Palmprint 251 Comparative Study of ROI Extraction of Palmprint 1 Milind E. Rane, 2 Umesh S Bhadade 1,2 SSBT COE&T, North Maharashtra University Jalgaon, India Abstract - The Palmprint region segmentation is an important

More information

Improvement of SURF Feature Image Registration Algorithm Based on Cluster Analysis

Improvement of SURF Feature Image Registration Algorithm Based on Cluster Analysis Sensors & Transducers 2014 by IFSA Publishing, S. L. http://www.sensorsportal.com Improvement of SURF Feature Image Registration Algorithm Based on Cluster Analysis 1 Xulin LONG, 1,* Qiang CHEN, 2 Xiaoya

More information

Image Registration. Prof. Dr. Lucas Ferrari de Oliveira UFPR Informatics Department

Image Registration. Prof. Dr. Lucas Ferrari de Oliveira UFPR Informatics Department Image Registration Prof. Dr. Lucas Ferrari de Oliveira UFPR Informatics Department Introduction Visualize objects inside the human body Advances in CS methods to diagnosis, treatment planning and medical

More information

Non-rigid Image Registration

Non-rigid Image Registration Overview Non-rigid Image Registration Introduction to image registration - he goal of image registration - Motivation for medical image registration - Classification of image registration - Nonrigid registration

More information

CS 223B Computer Vision Problem Set 3

CS 223B Computer Vision Problem Set 3 CS 223B Computer Vision Problem Set 3 Due: Feb. 22 nd, 2011 1 Probabilistic Recursion for Tracking In this problem you will derive a method for tracking a point of interest through a sequence of images.

More information

Matching. Compare region of image to region of image. Today, simplest kind of matching. Intensities similar.

Matching. Compare region of image to region of image. Today, simplest kind of matching. Intensities similar. Matching Compare region of image to region of image. We talked about this for stereo. Important for motion. Epipolar constraint unknown. But motion small. Recognition Find object in image. Recognize object.

More information

SIFT: SCALE INVARIANT FEATURE TRANSFORM SURF: SPEEDED UP ROBUST FEATURES BASHAR ALSADIK EOS DEPT. TOPMAP M13 3D GEOINFORMATION FROM IMAGES 2014

SIFT: SCALE INVARIANT FEATURE TRANSFORM SURF: SPEEDED UP ROBUST FEATURES BASHAR ALSADIK EOS DEPT. TOPMAP M13 3D GEOINFORMATION FROM IMAGES 2014 SIFT: SCALE INVARIANT FEATURE TRANSFORM SURF: SPEEDED UP ROBUST FEATURES BASHAR ALSADIK EOS DEPT. TOPMAP M13 3D GEOINFORMATION FROM IMAGES 2014 SIFT SIFT: Scale Invariant Feature Transform; transform image

More information

Image Coding with Active Appearance Models

Image Coding with Active Appearance Models Image Coding with Active Appearance Models Simon Baker, Iain Matthews, and Jeff Schneider CMU-RI-TR-03-13 The Robotics Institute Carnegie Mellon University Abstract Image coding is the task of representing

More information

Research on Evaluation Method of Video Stabilization

Research on Evaluation Method of Video Stabilization International Conference on Advanced Material Science and Environmental Engineering (AMSEE 216) Research on Evaluation Method of Video Stabilization Bin Chen, Jianjun Zhao and i Wang Weapon Science and

More information

arxiv: v1 [cs.cv] 2 May 2016

arxiv: v1 [cs.cv] 2 May 2016 16-811 Math Fundamentals for Robotics Comparison of Optimization Methods in Optical Flow Estimation Final Report, Fall 2015 arxiv:1605.00572v1 [cs.cv] 2 May 2016 Contents Noranart Vesdapunt Master of Computer

More information

Liver Image Mosaicing System Based on Scale Invariant Feature Transform and Point Set Matching Method

Liver Image Mosaicing System Based on Scale Invariant Feature Transform and Point Set Matching Method Send Orders for Reprints to reprints@benthamscience.ae The Open Cybernetics & Systemics Journal, 014, 8, 147-151 147 Open Access Liver Image Mosaicing System Based on Scale Invariant Feature Transform

More information

Feature descriptors and matching

Feature descriptors and matching Feature descriptors and matching Detections at multiple scales Invariance of MOPS Intensity Scale Rotation Color and Lighting Out-of-plane rotation Out-of-plane rotation Better representation than color:

More information

Hybrid Spline-based Multimodal Registration using a Local Measure for Mutual Information

Hybrid Spline-based Multimodal Registration using a Local Measure for Mutual Information Hybrid Spline-based Multimodal Registration using a Local Measure for Mutual Information Andreas Biesdorf 1, Stefan Wörz 1, Hans-Jürgen Kaiser 2, Karl Rohr 1 1 University of Heidelberg, BIOQUANT, IPMB,

More information

Face Hallucination Based on Eigentransformation Learning

Face Hallucination Based on Eigentransformation Learning Advanced Science and Technology etters, pp.32-37 http://dx.doi.org/10.14257/astl.2016. Face allucination Based on Eigentransformation earning Guohua Zou School of software, East China University of Technology,

More information

Advanced Image Reconstruction Methods for Photoacoustic Tomography

Advanced Image Reconstruction Methods for Photoacoustic Tomography Advanced Image Reconstruction Methods for Photoacoustic Tomography Mark A. Anastasio, Kun Wang, and Robert Schoonover Department of Biomedical Engineering Washington University in St. Louis 1 Outline Photoacoustic/thermoacoustic

More information

Open Access Moving Target Tracking Algorithm Based on Improved Optical Flow Technology

Open Access Moving Target Tracking Algorithm Based on Improved Optical Flow Technology Send Orders for Reprints to reprints@benthamscience.ae The Open Automation and Control Systems Journal, 015, 7, 1387-139 1387 Open Access Moving Target Tracking Algorithm Based on Improved Optical Flow

More information

Sketchable Histograms of Oriented Gradients for Object Detection

Sketchable Histograms of Oriented Gradients for Object Detection Sketchable Histograms of Oriented Gradients for Object Detection No Author Given No Institute Given Abstract. In this paper we investigate a new representation approach for visual object recognition. The

More information

EE795: Computer Vision and Intelligent Systems

EE795: Computer Vision and Intelligent Systems EE795: Computer Vision and Intelligent Systems Spring 2012 TTh 17:30-18:45 FDH 204 Lecture 11 140311 http://www.ee.unlv.edu/~b1morris/ecg795/ 2 Outline Motion Analysis Motivation Differential Motion Optical

More information

Object detection using non-redundant local Binary Patterns

Object detection using non-redundant local Binary Patterns University of Wollongong Research Online Faculty of Informatics - Papers (Archive) Faculty of Engineering and Information Sciences 2010 Object detection using non-redundant local Binary Patterns Duc Thanh

More information

The organization of the human cerebral cortex estimated by intrinsic functional connectivity

The organization of the human cerebral cortex estimated by intrinsic functional connectivity 1 The organization of the human cerebral cortex estimated by intrinsic functional connectivity Journal: Journal of Neurophysiology Author: B. T. Thomas Yeo, et al Link: https://www.ncbi.nlm.nih.gov/pubmed/21653723

More information

EE795: Computer Vision and Intelligent Systems

EE795: Computer Vision and Intelligent Systems EE795: Computer Vision and Intelligent Systems Spring 2012 TTh 17:30-18:45 FDH 204 Lecture 10 130221 http://www.ee.unlv.edu/~b1morris/ecg795/ 2 Outline Review Canny Edge Detector Hough Transform Feature-Based

More information

Semantic Context Forests for Learning- Based Knee Cartilage Segmentation in 3D MR Images

Semantic Context Forests for Learning- Based Knee Cartilage Segmentation in 3D MR Images Semantic Context Forests for Learning- Based Knee Cartilage Segmentation in 3D MR Images MICCAI 2013: Workshop on Medical Computer Vision Authors: Quan Wang, Dijia Wu, Le Lu, Meizhu Liu, Kim L. Boyer,

More information

Nonrigid Registration with Adaptive, Content-Based Filtering of the Deformation Field

Nonrigid Registration with Adaptive, Content-Based Filtering of the Deformation Field Nonrigid Registration with Adaptive, Content-Based Filtering of the Deformation Field Marius Staring*, Stefan Klein and Josien P.W. Pluim Image Sciences Institute, University Medical Center Utrecht, P.O.

More information

Spectral Log-Demons Diffeomorphic Image Registration with Very Large Deformations

Spectral Log-Demons Diffeomorphic Image Registration with Very Large Deformations Spectral Log-Demons Diffeomorphic Image Registration with Very Large Deformations Herve Lombaert Leo Grady Xavier Pennec Nicholas Ayache Farida Cheriet Abstract This paper presents a new framework for

More information

EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm

EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm Group 1: Mina A. Makar Stanford University mamakar@stanford.edu Abstract In this report, we investigate the application of the Scale-Invariant

More information

Motion Estimation for Video Coding Standards

Motion Estimation for Video Coding Standards Motion Estimation for Video Coding Standards Prof. Ja-Ling Wu Department of Computer Science and Information Engineering National Taiwan University Introduction of Motion Estimation The goal of video compression

More information

Face Alignment Under Various Poses and Expressions

Face Alignment Under Various Poses and Expressions Face Alignment Under Various Poses and Expressions Shengjun Xin and Haizhou Ai Computer Science and Technology Department, Tsinghua University, Beijing 100084, China ahz@mail.tsinghua.edu.cn Abstract.

More information

A New Technique of Extraction of Edge Detection Using Digital Image Processing

A New Technique of Extraction of Edge Detection Using Digital Image Processing International OPEN ACCESS Journal Of Modern Engineering Research (IJMER) A New Technique of Extraction of Edge Detection Using Digital Image Processing Balaji S.C.K 1 1, Asst Professor S.V.I.T Abstract:

More information

Fast Image Matching Using Multi-level Texture Descriptor

Fast Image Matching Using Multi-level Texture Descriptor Fast Image Matching Using Multi-level Texture Descriptor Hui-Fuang Ng *, Chih-Yang Lin #, and Tatenda Muindisi * Department of Computer Science, Universiti Tunku Abdul Rahman, Malaysia. E-mail: nghf@utar.edu.my

More information

Correspondence. CS 468 Geometry Processing Algorithms. Maks Ovsjanikov

Correspondence. CS 468 Geometry Processing Algorithms. Maks Ovsjanikov Shape Matching & Correspondence CS 468 Geometry Processing Algorithms Maks Ovsjanikov Wednesday, October 27 th 2010 Overall Goal Given two shapes, find correspondences between them. Overall Goal Given

More information

An Angle Estimation to Landmarks for Autonomous Satellite Navigation

An Angle Estimation to Landmarks for Autonomous Satellite Navigation 5th International Conference on Environment, Materials, Chemistry and Power Electronics (EMCPE 2016) An Angle Estimation to Landmarks for Autonomous Satellite Navigation Qing XUE a, Hongwen YANG, Jian

More information

Car Detecting Method using high Resolution images

Car Detecting Method using high Resolution images Car Detecting Method using high Resolution images Swapnil R. Dhawad Department of Electronics and Telecommunication Engineering JSPM s Rajarshi Shahu College of Engineering, Savitribai Phule Pune University,

More information

HOUGH TRANSFORM CS 6350 C V

HOUGH TRANSFORM CS 6350 C V HOUGH TRANSFORM CS 6350 C V HOUGH TRANSFORM The problem: Given a set of points in 2-D, find if a sub-set of these points, fall on a LINE. Hough Transform One powerful global method for detecting edges

More information

Illumination-Robust Face Recognition based on Gabor Feature Face Intrinsic Identity PCA Model

Illumination-Robust Face Recognition based on Gabor Feature Face Intrinsic Identity PCA Model Illumination-Robust Face Recognition based on Gabor Feature Face Intrinsic Identity PCA Model TAE IN SEOL*, SUN-TAE CHUNG*, SUNHO KI**, SEONGWON CHO**, YUN-KWANG HONG*** *School of Electronic Engineering

More information

Spectral Log-Demons: Diffeomorphic Image Registration with Very Large Deformations

Spectral Log-Demons: Diffeomorphic Image Registration with Very Large Deformations Spectral Log-Demons: Diffeomorphic Image Registration with Very Large Deformations Herve Lombaert, Leo Grady, Xavier Pennec, Nicholas Ayache, Farida Cheriet To cite this version: Herve Lombaert, Leo Grady,

More information

Object Recognition Algorithms for Computer Vision System: A Survey

Object Recognition Algorithms for Computer Vision System: A Survey Volume 117 No. 21 2017, 69-74 ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version) url: http://www.ijpam.eu ijpam.eu Object Recognition Algorithms for Computer Vision System: A Survey Anu

More information

Ensemble registration: Combining groupwise registration and segmentation

Ensemble registration: Combining groupwise registration and segmentation PURWANI, COOTES, TWINING: ENSEMBLE REGISTRATION 1 Ensemble registration: Combining groupwise registration and segmentation Sri Purwani 1,2 sri.purwani@postgrad.manchester.ac.uk Tim Cootes 1 t.cootes@manchester.ac.uk

More information

Image Segmentation Based on Watershed and Edge Detection Techniques

Image Segmentation Based on Watershed and Edge Detection Techniques 0 The International Arab Journal of Information Technology, Vol., No., April 00 Image Segmentation Based on Watershed and Edge Detection Techniques Nassir Salman Computer Science Department, Zarqa Private

More information

A NEW ROBUST IMAGE WATERMARKING SCHEME BASED ON DWT WITH SVD

A NEW ROBUST IMAGE WATERMARKING SCHEME BASED ON DWT WITH SVD A NEW ROBUST IMAGE WATERMARKING SCHEME BASED ON WITH S.Shanmugaprabha PG Scholar, Dept of Computer Science & Engineering VMKV Engineering College, Salem India N.Malmurugan Director Sri Ranganathar Institute

More information

Moving Object Segmentation Method Based on Motion Information Classification by X-means and Spatial Region Segmentation

Moving Object Segmentation Method Based on Motion Information Classification by X-means and Spatial Region Segmentation IJCSNS International Journal of Computer Science and Network Security, VOL.13 No.11, November 2013 1 Moving Object Segmentation Method Based on Motion Information Classification by X-means and Spatial

More information