Computer Vision – Overview Hanyang University Jong-Il Park Digital Image Processing Sampling & Quantization Image Enhancement Image Restoration Image Coding( or Image Data Compression) Image Understanding( or Computer Vision) Department of Computer Science and Engineering, Hanyang University Image formation Department of Computer Science and Engineering, Hanyang University Sampling & Quantization Department of Computer Science and Engineering, Hanyang University Department of Computer Science and Engineering, Hanyang University Subsampling & aliasing aliasing Department of Computer Science and Engineering, Hanyang University Sub-Nyquist sampling Circular Zone Plate Org. H ½, v ½ aliasing Department of Computer Science and Engineering, Hanyang University Image Enhancement Goal to accentuate certain image features for subsequent analysis or for image display Input : image Output : image Department of Computer Science and Engineering, Hanyang University Department of Computer Science and Engineering, Hanyang University Department of Computer Science and Engineering, Hanyang University Image Enhancement Techniques contrast/edge enhancement histogram equalization pseudo coloring noise filtering edge sharpening smoothing Applications processing of remote-sensed image via satellite radar, SAR, Ultrasonic image processing Department of Computer Science and Engineering, Hanyang University Image Restoration Goal to remove or minimize known/unknown degradations in image Department of Computer Science and Engineering, Hanyang University Department of Computer Science and Engineering, Hanyang University Blur model n ( x, y ) g ( x, y ) h( x, y; , ) f ( , )dd n( x, y ) or g Hf n ; discrete model estimate or recover f(x,y) from g(x,y) deconvolution problem Department of Computer Science and Engineering, Hanyang University Enhancement by integration Department of Computer Science and Engineering, Hanyang University Image Restoration Techniques deblurring noise filtering correction of geometric distortion inverse filtering Least mean square(Wiener) filtering Applications remote-sensed image processing noise cancellation Department of Computer Science and Engineering, Hanyang University Image Data Compression Goal to reduce the amount of data required to represent images Input : image Output : bit-stream data “010100101100110101001 . . . .” moving image: 7 1,0001,0008(bits/sample)330 frame/sec = 7210 bit/sec Department of Computer Science and Engineering, Hanyang University 2-D Image Compression, JPEG Image quality by JPEG (a) 27.9 dB at 0.125 bpp (b) 31.5 dB at 0.25 bpp (c) 34.8 dB at 0.5 bpp Department of Computer Science and Engineering, Hanyang University Department of Computer Science and Engineering, Hanyang University Department of Computer Science and Engineering, Hanyang University 3D Mesh Compression Model Bunny # of Vertex # of Triangle Original (byte) 34835 69473 2,879,786 Compression ratio Compressed (byte) 12 bpc 10 bpc 8 bpc 57,952 37,642 24,842 50 : 1 77 : 1 116 : 1 Original ‘Bunny’ 12 bpc 10 bpc 8 bpc Department of Computer Science and Engineering, Hanyang University Image Data Compression Techniques Error-free coding( or lossless coding) Lossy compression Image Compression Standard JPEG, H.261, H.263, MPEG-1,2,4 etc Applications Transmission teleconferencing ,TV system, remote sensing via satellite Storage VOD(video on demand), Video CD, DVD(digital video disk), medical imaging, educational and business documents Department of Computer Science and Engineering, Hanyang University Image Coding Standards Standard Application Bit rate JPEG Still-image Variable H.261 Visual telephony and video conferencing p64 kbps H.263 < 64 kbps MPEG-1 Full motion video on digital storage media 1.5 Mbps MPEG-2 Higher resolution video than MPEG-1 2 Mbps HDTV Terrestrial broad cast 20 Mbps Department of Computer Science and Engineering, Hanyang University MPEG MPEG-1 1992 Video CD Includes Audio Layer III (MP3) MPEG-4 Part 1: 1998 Part 2: 1999 (addition still under development) Audio visual object Mobile telephony (IMT-2000) Interactivity Content –Based Interactivity Department of Computer Science and Engineering, Hanyang University MPEG-7 2001 (addition still under development) Content Description Interface Metadata language and schemes for search & retrieval Department of Computer Science and Engineering, Hanyang University Content-based Image Retrieval Image retrieval Find similar images from image database Used features Color Texture Shape Query Retrieved Department of Computer Science and Engineering, Hanyang University Query Image Image Data List Retrieval Results Texture Option Department of Computer Science and Engineering, Hanyang University Content-based Video Retrieval Video retrieval Scene change detection and key frame extraction Extracted key frame from a movie clip Department of Computer Science and Engineering, Hanyang University Content-based Video Retrieval Video retrieval Automatic indexing Index from a news sequence Department of Computer Science and Engineering, Hanyang University Media Framework Tech. MPEG-21 All electronic creation, delivery and trade of digital multimedia content Transparent usage of various content types on various devices connected through various network Department of Computer Science and Engineering, Hanyang University Watermarking the process of embedding information into a host signal without changing perceived characteristics of the host Key Image/ Document Watermark Information Watermark Embedding Key * Filtering * Lossy Compression * Forgery * Cropping & Scaling * Geometric Distortion Watermark Detection Decoded Watermark * Copy Protection * Fingerprinting * Authentification Watermark Generation Department of Computer Science and Engineering, Hanyang University Image Understanding Goal to interpret or describe the meaning contained in the image Input : image Output : interpretation(description) “KAISION” “circle” Department of Computer Science and Engineering, Hanyang University Department of Computer Science and Engineering, Hanyang University Department of Computer Science and Engineering, Hanyang University Stereo Vision (disparity map) 기준 영상(좌) 기준 영상(우) 변위 참 값(붉은색: 가리워짐) 구한 변위 지도(검은색: 가리워짐) Department of Computer Science and Engineering, Hanyang University Scene Matching (1) Model image Model Scene matching result Input scene Department of Computer Science and Engineering, Hanyang University Scene Matching (2) Model Scene matching result Input image Edge image Department of Computer Science and Engineering, Hanyang University Fruit Classifier Apple Orange Water Mellon – 크기, 반지름 Department of Computer Science and Engineering, Hanyang University – color Department of Computer Science and Engineering, Hanyang University Edge Detection (a) Lena OriginalImage(256 256) (b) CannyOperator( 2, low 5, high 10) Department of Computer Science and Engineering, Hanyang University Segmentation (a) Aerial Original Image (b) Segmented Image Department of Computer Science and Engineering, Hanyang University Department of Computer Science and Engineering, Hanyang University There are a lot of interesting topics to explore in image processing and computer vision. Department of Computer Science and Engineering, Hanyang University