Computer Vision. I-Chen Lin, Assistant Professor Dept. of CS, National Chiao Tung University

Size: px

Start display at page:

Download "Computer Vision. I-Chen Lin, Assistant Professor Dept. of CS, National Chiao Tung University"

Lesley Harmon
5 years ago
Views:

1 Computer Vision I-Chen Lin, Assistant Professor Dept. of CS, National Chiao Tung University

2 About the course Course title: Computer Vision Lectures: EC016, 10:10~12:00(Tues.); 15:30~16:20(Thurs.) Pre-requisites: Computer programming skills in C/C++. Moderate levels to handle data structures. (optional) related courses: e.g. introduction to computer graphics, image processing, pattern recognition. Teacher: I-Chen Lin ( 林奕成 ), Assistant Professor ichenlin@cs.nctu.edu.tw Office: EC 704 ( 工程三館 )

3 About the course (cont.) TAs: 蔡明翰卓孟虹 Office: EC 237 (EC229b) Phone ext: (56676) Course web page: Textbook David A. Forsyth and Jean Ponce, Computer Vision: A Modern Approach, Prentice Hall, New Jersey. (1 st or 2 nd ed.) Reference book: Richard Hartley and A. Zisserman, Multiple View Geometry in Computer Vision 2nd Ed., Cambridge University Press, 2004.

4 About the course (cont.) References IEEE Trans. Pattern Analysis and Machine Intelligence (PAMI). Intl. J. Computer Vision (IJCV). Proc. Intl. Conf. Computer Vision (ICCV). Proc. Intl. Conf. Computer Vision and Pattern Recognition (CVPR). Proc. Euro. Conf. Computer Vision (ECCV). ACM Trans. Graphics/SIGGRAPH/SIGGRAPH Asia/ Some contents are borrowed from the reference lecture notes: Prof. D.A. Forsyth, Computer Vision, UIUC. Prof. T. Darrell, Computer Vision and Applications, MIT. Prof. J. Rehg, Computer Vision, Georgia Inst. of Tech. Prof. D. Lowe, Computer Vision, UBC, CA. Prof. C.F. Chang, Image-based rendering, NTHU/NTNU. Prof. S. Seitz and P.Heckbert, Image-based modeling and rendering, CMU.

5 What s computer vision? The science of extracting information about the world from images. How to discover from images what is present in the world, where things are, what actions are taking place. (Marr 1982) One of the most challenging mysteries in Computer Science! Closely related fields: Image processing Artificial intelligence and machine Learning Computer Graphics

6 Vision and related fields Outputs descriptions images Input descriptions images Computer Vision & Pattern Recognition IBMR Computer Graphics Image Processing

7 Computer Graphics Figures from SIGGRAPH 99 Course Notes IBMR

8 Computer Vision Figures from SIGGRAPH 99 Course Notes IBMR

9 Knowing the scene Those common (or even trivial) abilities for humans can be quite difficult for computers.

10 Vision is fundamentally Ill-Posed There are an infinite number of possible scenes that could result in the pixels in a captured image. Geometry Photometry Camera Sensor plane Alternate Surface Surface Figure from J. Rehg s lecture note: Computer Vision, Georgia Inst. Tech.

11 Monocular, static cues Human perception makes use of prior knowledge about the shape and: Relative size Occlusion Perspective Linear Aerial

12 Monocular, static cues Lighting Shadow Texture gradients

13 Monocular, dynamic cues Motion parallax

14 Binocular cues wickelgren/psyc110/stereopsis.jpg

15 Stereo triangulation Left view Right view The estimated depth image (map) Synthetic view with texture mapping

16 Syllabus Perspective, lens and camera Canon Power Shot A95 Figure from

17 Syllabus (cont.) Radiometry Illumination and reflectance models Kettle, Mike Miller, POV-Ray R N L

18 Syllabus (cont.) Color

19 Syllabus (cont.) Feature extraction: edge, corner, SIFT, etc. D. Frolova, D. Simakov, Slides of Matching with Invariant Features.

20 Syllabus (cont.) Image matching and panorama M. Brown and D.G. Lowe, Automatic Panoramic Image Stitching using Invariant Features, IJCV 2007

21 Syllabus (cont.) Clustering and segmentation Mean-shift clustering Graph-cut segmentation

22 Syllabus (cont.) Structure from motion Shaded model Static scene 3D model reconstructed by Luc Van Gool et al.

23 Syllabus (cont.) Advanced topics S.Seitz et al. View Morphing, SIGGRAPH 96 P.Tan et al., Image-based Tree Modeling,SIGGRAPH 07 Eigenface, M. Glencross, et al., A Perceptually Validated Model for Surface Depth Hallucination, Proc. SIGGRAPH 08.

24 About the course (cont.) Grades: (temporarily) Exams (25~30%) Homework (40%) Photometric stereo, segmentation or keypoint-related Term project (30~35%) 1~3 members per group Paper and proposal presentation Demo & final presentation. Class participation (bonus)

25 The schedule Proposal Presentation Demo & presentation Course beginning Homework lectures Exam

I Chen Lin, Assistant Professor Dept. of CS, National Chiao Tung University. Computer Vision: 6. Texture

I Chen Lin, Assistant Professor Dept. of CS, National Chiao Tung University Computer Vision: 6. Texture Objective Key issue: How do we represent texture? Topics: Texture analysis Texture synthesis Shape