Digital Video And Compression
- 1. Digital Video and Compression Robert Burk C.S. 41 Spring 2002
- 2. Overview Background Digital video Why digital video Video compression
- 3. Background First “transmission” of visual images 1885 by George Carey in Boston Image focused on a photo cell array Lamp array displayed the image Conventional television Developed in the 1930’s Technology limitations of the time Human vision characteristics Television today based on this early work
- 4. Background NTSC Adopted in 1941 525 lines, interlaced scan Aspect ratio of 4:3 Analog format Color added to standard in 1953 Analog transmission “compression” Vestigial sideband used to reduce bandwidth Video and audio in a 6 MHz band
- 5. Why Digital Video? Started with tape recording Record, edit , play, record, edit, play causes losses in analog Digital recording not nearly as lossy - better s/n ratio Migrated to other video products in the production chain – avoided d/a and a/d conversions
- 6. Digital Video DTV based on ATSC Timeline May 1993 – Grand Alliance Nov 1995 – committee submitted recommendation to ATSC Dec 1996 – FCC ruling to accept the standard and make the requirements The standard primarily supports HDTV – or alternatively several SDTV signals within the existing 6 MHz bandwidth
- 7. Advantages of DTV Signal quality No ghosting, no snow, aircraft flutter Enhanced services HDTV bandwidth – bigger, clearer picture, 16:9 aspect ratio 5.1 Channels of CD quality sound Multiple signal sources on the same channel (more program choices) Other digital services – integration with computer industry
- 8. OSI 7-Layer Model of ATSC
- 9. Data Rates of DTV Audio and video sources Up to 1.5 Gbps A variety of standards for both the picture and the sound 1080i 720p Dolby digital (AC-3) sound Which one is standard? None of them! Sounds good, but how do we get that much data in 6 MHz?
- 10. Compression! It’s simple – compression! Data reduction of 60:1 and greater is required for the video in ATV systems A number of techniques are used, including both lossless and lossy compression
- 12. Discrete Cosine Transform Statistical data redundancy or spatial redundancy – exploiting the fact that identical data need not be repeated Psycho visual redundancy – throwing out the things that you would not notice if they were removed DCT coefficient tables used in the encoding and decoding scheme Similar to the art form know as Photomosaic
- 13. Variable Length Coding Also known as Huffman coding and entropy coding Based on the probability of identical values appearing Assigns shortest codes to values with highest probability – longest codes to lowest probability values
- 14. Run-Length Coding Relies on the repetition of the same data No need to send the same data multiple times – just send how many times to repeat the same value Techniques such as DCT tends to create long strings of the same values
- 15. Blanking Removal In NTSC, horizontal and vertical blanking pulses provided to provide timing – relatively long pulses The timing information is still needed for the proper display of the picture Replace this information with shorter synchronization data strings specific to the application
- 16. Sub-sampling Reduces the numbers of samples taken Very effective data reduction method, but loss of picture resolution and “aliasing” can degrade picture quality Therefore, sub-sampling not applied to luminance data – only to chrominance data Referred to as the 4:2:0 or 4:1:1 signal structure (Luminance: C B :C R )
- 17. Differential Pulse Code Modulation DPCM is a predictive encoding scheme Transmits the sample-to-sample differences rather than the full value The difference is added to the current decoded value at the receiving end to generate a reconstructed sample value
- 18. Quantization Representation of value by the closest approximation from a set of defined values (reduced amplitude resolution) E.g., if we code a signal to 8 bits, we determine which of 256 possible value is closest to the value to be encoded and use that value In video, the precision required depends on the viewing conditions – aim for 100:1 contrast ratio
- 19. Other Techniques Motion JPEG Temporal compression Macro Blocks and slices Object-based coding Packetized Elementary Stream (PES) Audio compression
- 20. In Summary … Today, though we are still watching analog video, it is all been digital Compression techniques will allow the whole delivery chain to go digital More choices will be available both in terms of what is available, as well as what you do with it and how you process it The question that still remains to be answered … When will all of this be available at a price we can afford?