Authorization In HSV-DWT-DCT - For Depth-Image-Based Rendering 3D Images

Transcription

1 Authorization In HSV-DWT-DCT - For Depth-Image-Based Rendering 3D Images Arun K A & Jeno poul P Department of electronics and communication PSN college of engineering and technologies connecttoarunka@gmail.com,jenopaul1@rediffmail.com Abstract - The content protection for image-based 3D data is getting more importance with the advance of low cost 3D display devices. The depth-image-based rendering (DIBR) 3D image is one of the image-based 3D data which consists of the centre image and the depth image generated by the content provider. The left-eye image and the right-eye image are rendered from the centre image and the depth image at the content consumer side. The blind watermarking for DIBR 3D image is rarely studied in the literature. In this project, a novel blind multiple watermarking scheme is proposed to deal with the content protection problem of DIBR 3D images. Besides the usual requirement of mutual orthogonally among reference patterns for multiple water mark embedding, we found that proper embedding order plays an even more important role in watermarking the DIBR 3D images. Experimental results show that the proposed scheme is robust against the JPEG compression and noise adding attacks. More interestingly, it is found that the proposed watermarking can also tolerate large range variations of the depth image during rendering. Keywords Blind watermarking, RGB-HSV, Discrete wavelet transform, Discrete Cosine transform, watermark embedding Algorithm, Rendering process. I. INTRODUCTION With the emerging of low cost 3D display devices, different kinds of 3D applications and the amount of 3D content are booming up, recently. Effective digital watermarking systems for 3D content protection will play an important role for developing and promoting the 3D entertainment industry. The 3D watermarking systems can be classified on the basis of their embedding space and detecting space as: 3D/3D, 3D/2D and 2D/2D. The first kind of watermarking system embeds the watermark in a 3D (geometry or texture) model and detects the watermark also in a 3D space. The second one extracts the watermark in a 2D representation (acquired by projection) of the 3D data. The last one deals with the content protection of imagebased 3D data representation in which all the embedding and detection operations are applied to 2D images directly. Since 3D data are usually distributed in imagebased representations, the 2D/2D watermarking scheme becomes one of the most popular approaches for preventing the increasing amount of 3D content from being distributed and used illegally. Human visual system (HVS) perceives the depth illusion of a 3D image from two different viewpoint images, one from the left-eye and another from the right-eye. Generally, there are two major image-based representations for 3D images. The first one is the socalled stereo image, where the left-eye image and the right-eye image are recorded directly. The second one is the depth-image-based rendering (DIBR) 3D image that records the centre image and the depth image. Due to the specific data dependency characteristics of 3D images, all the works need the original image or the watermarked image without noise during watermark detection. It is well-known that a blind watermarking system is more preferable because the original data is not needed during watermark detection, which coincides with the requirement of most 3D application scenarios. The proposed method aims to develop a light way watermark scheme, suitable for a web-crawler watermark detector, in which the original data or the corresponding depth image is excluded during the watermark detection. In other words, a blind watermarking system for DIBR 3D images is proposed and investigated, in this work. As prescribed, the virtual left-eye and right-eye images are obtained in the receiver side by rendering the centre and the depth images. The watermark embedded in the centre image has to be kept on the virtual left-eye and right-eye images after the rendering.. 131

2 Fig. 1 shows the digital communication model of a DIBR 3D image system, in which the receiver side performs DIBR operation to generate the left-eye and the right-eye images. During the transmission, both the depth image and the center image will be disturbed by the channel noise. Fig. 2 illustrates the illegal redistribution performed in the content consumer side where the depth image may not be distributed, concurrently. If the left-eye and the right-eye images are illegally re-distributed, the depth image may not be available for watermark extraction. Therefore, a blind watermarking scheme is a nature choice for protecting DIBR 3D images. Without the aid of depth image, the rendering operation can be viewed as a kind of local geometric modification, which is an effective attack for many blind watermarking schemes. There are lots of robust blind watermarking schemes for 2D images that can resist geometric modification attacks.. Fig 1.Digital communication model of a DIBR 3D image Fig. 2. Illegal content re-distribution in the consumer side. 132

3 However, the color plus depth 3D images introduce additional rendering induced noise, which diminishes the robustness of these watermarking schemes. Therefore, finding an appropriate approach to improve the robustness of watermarking schemes for DIBR 3D images is the main target of this work. Interestingly, we found that the rendering induced noise can be eliminated by using a multiple watermark scheme. The reason for the effectiveness of this remedy can be summarized as follows. Although the rendering operation is one kind of local geometric modification; fortunately, we have the full knowledge of the rendering operation in the watermark embedding phase. That means we may reduce its negative effect by utilizing our full embedded in the center image should be kept both on the understanding of the rendering operation. The proposed blind watermark scheme is illustrated in Fig. 3, where the watermark left-eye and the right-eye images after rendering. This paper is organized as follows. Section II reviews the basic DIBR operations. Section III presents the proposed watermarking scheme for DIBR 3D images. Experimental results are demonstrated in Sections IV and V concludes this work. II. DIBR 3D IMAGE SYSTEMS The depth image indicates the depth values of the objects in a 3D scene. DIBR system warps the center image according to the depth values to synthesize the left-eye and the right-eye images. A. Rendering Operation The rendering operation for synthesizing the lefteye and/or the right-eye images is performed by warping the center n image pixel-wisely. Figure 3 shows an example of center image, depth image and the corresponding mapping between intensity and depth value.. Center image, (b) Depth image, and (c) The mapping between intensity and depth value Fig 3. Depth Image Representation..The depth value ranges from Znear to Zfar, where Znear is the nearest clipping plane and Zfar is the farthest clipping plane in the 3D scene. The depth value maps the intensity value linearly into the range [Zfar, Znear] (which will be [0,255] in an 8-bit depth image). The rendering operation warps the pixel location according to the depth value by the following formula: xl=xc+(tx/2 f/z) (1) xr=xc-(tx/2 f/z) (2) Where Xc is the -coordinate of a pixel in the center image, XL/XR is the corresponding -coordinate of the pixel in the left-eye/right-eye image, is the focal length of camera, tx is the baseline distance or camera distance, and Z represents the depth value of the current pixel in the center image. According to the visibility property of a 3D scene, the near object will occlude the object with a distant; therefore, in the rendering process, the pixels with the farthest depth value should be warped first. III. HOLE-FILLING PROBLEMS There is a well-known problem for DIBR 3D images: the left-eye or right-eye image may have holes due to sharp changes in the depth image. The existence of holes represents the situation that the 3D objects that are occluded in the center image but are visible in the left-eye or right-eye image. The Hole-Filling algorithm can simply be a linear interpolation done in the receiver side; of course, there are lots of studies dealt with the Hole-Filling problem by providing better synthesis quality according to the depth image. Figure 3.2 shows an example of the left-eye image with holes and the image with holes being filled by linear interpolation. In order to provide the flexibility for Hole-Filling algorithm selection in a DIBR 3D image system, this study takes only the number of holes into account for 133

4 designing watermarking scheme. Without loss of generality and for simplicity, this paper uses linear interpolation as the Hole- Filling algorithm in the following discussions. Of course, better quality of the reconstructed images could be expected if a more complicated Hole-Filling algorithm were used. been used by papermakers to identify their product, and also on postage stamps, currency, and other government documents to discourage counterfeiting. In France, they were introduced during World War II by the Vichy regime, and counterfeited by people such as Adolfo Kaminsky. The invention of the dandy roll revolutionized the watermark process and made it much easier for a company to watermark their paper. Cylinder Mould Process (a) Fig 4 Hole filling Left-eye image with holes and (b) the holefilled image by using linear interpolation IV. WATERMARK A watermark is a recognizable image or pattern in paper that appears as various shades of lightness/darkness when viewed by transmitted light (or when viewed by reflected light, atop a dark background), caused by thickness or density variations in the paper. There are two main ways of producing watermarks in paper; the dandy roll process, and the more complex cylinder mould process. Watermarks vary greatly in their visibility; while some are obvious on casual inspection, others require some study to pick out. Various aids have been developed, such as watermark fluid that wets the paper without damaging it. Watermarks are often used as security features of banknotes, passports, postage stamps, and other documents to prevent counterfeiting. A watermark is very useful in the examination of paper because it can be used for dating, identifying sizes, mill trademarks and locations, and the quality of a paper. Encoding an identifying code into digitized music, video, picture, or other file is known as a digital watermark. Dandy Roll Process A watermark is made by impressing a water-coated metal stamp or dandy roll onto the paper during manufacturing. These watermarks were first introduced in Bologna, Italy, in 1282; however the dandy roll was invented in 1826 by John Marshall. Watermarks have Another type of watermark is called the cylinder mould watermark. A shaded watermark, first used in 1848, incorporates tonal depth and creates a grey scale image. Instead of using a wire covering for the dandy roll, the shaded watermark is created by areas of relief on the roll's own surface. Once dry, the paper may then be rolled again to produce a watermark of even thickness but with varying density. The resulting watermark is generally much clearer and more detailed than those made by the Dandy Roll process, and as such Cylinder Mould Watermark Paper is the preferred type of watermarked paper for banknotes, passports, motor vehicle titles, and other documents where it is an important anti-counterfeiting measure. V. THE PROPOSED METHOD Like every coin, which has two sides, the digital media also has its own share of problems. The ease of accessibility of digital media and the simplicity of the digital systems has rendered the contents over the digital media highly insecure. Digital entities can be easily duplicated, manipulated, or even tampered. Thus the question of copyright is associated with a digital entity that faces a severe threat from hackers. Engineers have accepted this challenge in a gallant way. Many techniques have been devised to protect the copyright of digital entities. The technique of digital watermarking is one of the growing fields in which this problem of copyright has been addressed elegantly. The digital watermark is a secret code or image hidden inside the original image, so as to claim for the copyright of that image. Thus the process by which the copyright information is embedded invisibly inside the original entity, which is to be protected from the illegal replication and distribution, is known as Digital Watermarking. The success of the watermarking Scheme largely depends upon the choice of the watermark structure and insertion strategy. The two main constraints involved in the problem of watermarking are those of maintaining the robustness of the watermark information while keeping visual perception of the original image intact. 134

5 In watermarking the attacks on the original data can be classified as Intentional and unintentional. Unintentional attacks include the common signal processing such as low pass filtering, median filtering, digital to analog and analog to digital conversion, resampling and re-quantization, and common geometric distortions such as rotation, translation, cropping and scaling. Intentional attacks include collusion and forgery. Watermarking techniques can be classified in several ways. According to the need of the original image for watermark extraction or detection, watermarking is classified to non-blind and blind watermarking techniques, such as, require that the original image exists when detecting the watermark, whereas, blind techniques do not. Another way to classify watermarking is by how the watermark is inserted; in the spatial domain, or in the transform domain. Early watermarking schemes that were introduced, worked in the spatial domain, where the watermark is added by modifying pixel values of the host image. Some of the spatial domain watermarking approaches were based on the modification of the least significant bit (LSB) of an image based on the assumption that the LSB data are insignificant Generally, spatial domain watermarking is easy to implement from a computational point of view, but too fragile to resist numerous attacks. In order to have more promising techniques, researches were directed towards watermarking in the transform domain, where the watermark is not added to the image intensities, but to the values of its transform coefficients. Then to get the watermarked image, one should perform the transform inversely. Some of the transform based watermarking techniques used the Discrete Cosine Transform (DCT). The wavelet transform is another type of the transform domain. Wavelet based transform gained popularity recently since the property of multi resolution analysis that it provides. However, DWT has been used in digital image watermarking more frequently due to its excellent spatial localization and multi-resolution characteristics, which are similar to the theoretical model of human visual system. Further performance improvements in DWT based digital image watermarking algorithms could be obtained by combining DWT with DCT. The idea is based on the fact that combined transforms could compensate for the drawbacks of each other, resulting in effective watermarking. In this paper, we will describe a digital image watermarking algorithm based on combining two transforms; DWT and DCT. Watermarking is done by altering the wavelets coefficients of carefully selected DWT sub-bands, followed by the application of the DCT transform on the selected sub-bands. In order for a digital watermarking method to be effective it should be imperceptible, and robust to common image manipulations like compression, filtering, rotation, and scaling, cropping, collusion attacks among many other digital signal processing operations. In this project, we will describe a digital image watermarking algorithm based on combining two transforms; DWT and DCT. Watermarking is done by altering the wavelets coefficients of carefully selected DWT sub-bands, followed by the application of the DCT transform on the selected sub-bands. Watermarking is done altering the wavelet coefficient of selected sub bands, followed by the application of DCT transform on the selected sub bands. The rest of this paper is organized as follows. A. Watermark Embedding Procedure The main strength offered by transform domain techniques is that they can take advantage of special properties of alternate domains to address the limitations of spatial domain or to support additional features. We start the watermarking process by applying DWT to the host image, and afterwards performing the DCT to the selected DWT sub-bands. The agreement adopted by many DWT-based watermarking methods, is to embed the watermark in the middle frequency sub-bands HLX and LHX is better in perspective of imperceptibility and robustness. Consequently, HLX sub-bands in level three is chosen for Performing DCT on them. B. Algorithm Step1: Convert the Central image into HSV image Step2: Perform DWT on the resulting value component of the original image to decompose it into four nonoverlapping multi resolution coefficient sets: LL1, HL1, LH1 and HH1. Step3: Perform DWT again on two HL1 and LH1 subbands to get eight smaller sub-bands and choose four coefficient sets: HL12, LH12, HL22 and LH22. Step4: Perform DWT again on four sub-bands:hl12, LH12, HL22 and LH22 to get sixteen smaller Sub bands and choose four coefficient sets: HL13, LH13, HL23 and LH23. Step5: Divide four coefficient sets: HL13, LH13, HL23 and LH23 into 4 x 4 blocks. Step6: Perform DCT to each block in the chosen coefficient sets (HL13, LH13, HL23 and LH23) and form the result into single matrix. Step7:Re-formulate the grey-scale watermark image into a vector of zeros and ones. 135

6 Step8: Generate two uncorrelated pseudorandom sequences by a key. One sequence is used to embed the watermark bit 0 (PN_0) and the other sequence is used to embed the watermark bit 1 (PN_1). Number of elements in each of the two pseudorandom sequences must be equal to the number of mid-band elements of the DCT-transformed, DWT coefficient sets. Step9:Embedd the two pseudorandom sequences, PN_0 and PN_1,in the DCT transformed 4x4 blocks of the selected DWT coefficient sets of the host image. Embedding is not applied to all coefficients of the DCT block, but only to the mid-band DCT coefficients. If we donate X as the matrix of the mid band coefficients of the DCT transformed block, then embedding is done as follows: If the watermark bit is 0 then X '= X + α * PN_0 (14) If the watermark bit is 1 then X '= X + α * PN_1 (15) Step 10: Perform inverse DCT (IDCT) on each block after its mid-band coefficients have been modified to embed the watermark bits as described in the previous step. Step 11: Perform the inverse DWT (IDWT) on the DWT transformed image, including the modified coefficient sets, to produce the watermarked host image. Convert this HSV watermarked image to RGB watermarked image. C. Watermark Extraction Procedure The Combined DWT-DCT algorithm is a blind watermarking algorithm & in this the original host image is not required to extract watermark. Transformed 4x4 blocks of the selected DWT coefficient sets of the host image. Embedding is not applied to all coefficients of the DCT block, but only to the mid-band DCT. D. Algorithm Step1: Convert the watermarked image into HSV image Step2: Perform DWT on the value component of prefiltered watermarked image to decompose it into four non overlapping multi-resolution coefficient sets: LL1, HL1, LH1 and HH1. Step3: Perform DWT again on two sub-bands HL1 and LH1 to get eight smaller sub-bands and choose four coefficient sets: HL12, LH12, HL22 and LH22. Step4: Perform DWT again on four sub-bands:hl12, LH12, HL22 and LH22 to get sixteen smaller sub bands and choose four coefficient sets: HL13, LH13, HL23 and LH23. Step5: Divide four coefficient sets: HL13, LH13, HL23 and LH23 into 4 x 4 blocks. Step6: Perform DCT on each block in the chosen coefficient sets (HL13, LH13, HL23 and LH23) and form the result into single matrix. Step7: Regenerate the two pseudorandom sequences (PN_0 and PN_1) using the same key which used in the watermark embedding procedure. Step8: For each block in the coefficient sets: HL13, LH13, HL23 and LH23 calculate the correlation between the mid-band coefficients and the two generated pseudorandom sequences (PN_0 and PN_1). If the correlation with the PN_0 was higher than the correlation with PN_1, then the extracted watermark bit is considered 0, otherwise the extracted watermark is considered 1. Step9: The watermark is reconstructed using the extracted watermark bits, and compute the similarity between the original and extracted watermarks Figure 9 Block diagram of Extracting process 136

7 VI. EXPERIMENTAL RESULTS The proposed watermark extraction system can generate false positive errors that depend on and, mentioned in the watermark decoding process. In order to estimate the error rate, we select coefficients from ten randomly selected un watermarked images. The quantization error differences of the two unquantized coefficients in a pair are calculated by the same method in the watermark extraction process. The histograms of the quantization error and its estimated distribution are shown in Fig. 10. The quantization error difference is modeled with sum of Gaussian distribution as follows Figure 10. Histogram of the quantization error difference for the unquantized coefficients and its estimated model. To show the correctness of the theoretical probability error With a comparison to the experimental results, a BER experiment for unwatermarked images was conducted. Fig. 11 shows the histogram of BERs for the experimental result of unwatermarked images and probability density function Fig. 11. Comparison with theoretical false positive error and experimental false positive error. V. CONCLUSION The emerging 3D market inevitably produces copyright issues of the 3D content. In this work, I proposed blind watermarking scheme for DIBR 3D images with consideration of various scenarios, including a single virtual view being illegally distributed. In order to make the proposed watermarking robust to DIBR, 3 level DWT and after DCT are taken. Furthermore, pre-processing of depth images and baseline distance adjustment, which have not been issues in previous watermarking on DIBR 3D image, are considered to design the proposed watermarking method by exploiting the characteristics of the DWT-DCT. It is also robust to pre-processing of the depth image and baseline distance adjustment. Digital watermarking for 3D content is still in the early stages, and the standard of the DIBR is still being actively researched. In this situation, I will focus on extending my scheme to DIBR 3D videos and various standards of DIBR system. Also, improvement of the fidelity in terms of objective test and subjective test will be taken into consideration in future work. VI. REFERENCES [1] A. Chammem, M. Mitrea, and F. Prłteux, DWTbased stereoscopic image watermarking, in Proc. SPIE 7863, 2011, p [2] J.-L. Dugelay, S. Roche, C. Rey, and G. Doerr, Still-image watermarking robust to local geometric distortions, IEEE Trans. Image Process., vol. 15, no. 9, pp , Sep [3] C. Fehn, Depth-image-based rendering (dibr), compression, and tsransmission for a new approach on 3d-tv, SPIE Stereoscopic Displays Virtual Reality Syst. XI, vol. 5291, no. 1, pp , [4] E. Halici and A. Alatan, Watermarking for depth-image-based rendering, in IEEE Int. Conf. Image Process. (ICIP), 2009, Nov. 2009, pp [5] Hee-Dong Kim, Ji-Won Lee, Tae-Woo Oh, and Heung-Kyu Lee, Robust DT-CWT Watermarking for DIBR 3D Images, IEEE Trans Broadcasting, vol. 58, no. 4, December [6] C.-T. Hsu and J.-L. Wu, Hidden digital watermarks in images, IEEE Trans. Image Process., vol. 8, no. 1, pp , Jan [9] D.-C. Hwang, K.-H. Bae, M.-H. Lee, and E.-S. Kim, Real-time stereo image watermarking 137

8 using discrete cosine transform and adaptive disparity maps [7] A. Koz, C. Cigla, and A. Alatan, Free-view watermarking for freeview television, in IEEE Int. Conf. Image Process., Oct. 2006, pp [8] Y. H. Lin and J. L.Wu, A digital blind watermarking for depth-image based rendering 3D images, IEEE Trans. Broadcasting., vol. 57, no. 2, pp , Jun [9] P. J. Lee and Effendi, Nongeometric distortion smoothing approach for depth map preprocessing, IEEE Trans. Multimedia, vol. 13, no. 2, pp , April