1. Advanced JPEG Dinesh Ganotra

2. Organizations Involved Joint Photographic Experts Group ISO International Organization for Standards CCITT International Telegraph and Telephone Consultative Committee IEC International Electro-technical Commission

3. Recap

4. Contrast Sensitivity of the Eye ◦ At a 1 meter viewing distance an ideal digital image 40cm wide should have about 4,000 pixels. ◦ The contrast sensitivity curve for luminance is different from chrominance.

5. QUANTIZATION TABLE Luminance Quantization TableChrominance Quantization Table

6. ZERO SHIFT The 8-bit image data is zero-shifted by subtracting 128 from the data before the DCT is calculated. For 12-bit data the zero shift is accomplished by subtracting 2048. This reduces the internal precision requirement for DCT calculation.

7. DCT One multiplication and nine additions per coefficient × +++++++++

8. Entropy PCM DPCM Huffman Coding Arithmatic Coding 0 001 1011 100 101

9. DPCM lossless lossy DCT Sequential Progressive HierarchicalHuffmanArithmetic

10. Huffman Computationally Simpler Requires code tables to be known Arithmetic Higher Performance ( 10%) code book adapts dynamically to the data being coded.

11. Sequential codes the image in a single scan or passes through the data. The decoder necessarily follows the same order in decoding the data Progressive encodes the image in two or more scans or passes through the data, first coding an approximation of the entire image with each succeeding scan. decoder follows the same order in decoding, first reconstructing the approximating of the image, and then adding finer detail with each succeeding scan

12. 0 1 2 magnitude bit plane AC band 4 AC band 2 AC band 3 AC band 1 AC band 5 AC band 6 DC band 1 DC band 4 Progressive

13. Hierarchical progressive coding with increasing spatial resolution between progressive stages

14. Data Unit Y1Y1 Y2Y2 Y3Y3 Y4Y4 Y5Y5 Y6Y6 Y7Y7 Y8Y8 Cb 1 Cb 2 Cb 3 Cb 4 Cr 1 Cr 2 Cr 3 Cr 4 Vertically interleaved data units Y1Y1 Y2Y2 Y5Y5 Y6Y6 Y3Y3 Y4Y4 Y7Y7 Y8Y8 Cb 1 Cb 2 Cb 3 Cb 4 Cr 1 Cr 2 Cr 3 Cr 4 Horizontally interleaved data units

15. Non interleaved, horizontal, and vertical interleaved Scan1 : Y 1, Y 2, Y 3, Y 4, Y 5, Y 6, Y 7, Y 8 Scan2 : Cb 1, Cb 2, Cb 3, Cb 4 Scan3 : Cr 1, Cr 2, Cr 3, Cr 4 Scan1 : Y 1, Y 2, Cb 1, Cr 1, Y 3, Y 4, Cb 2, Cr 2, Y 5, Y 6, Cb 3, Cr 3, Y 7, Y 8, Cb 4, Cr 4 Scan1 : Y 1, Y 3, Cb 1, Cr 1,, Y 2, Y 4, Y 5, Y 7, Cb 3, Cr 3, Y 6, Y 8, Y 7, Y 8, Cb 4, Cr 4

16. Example FFD8 FFC9 00 11 08 00 80 00 80 03 01 11 00 02 11 00 03 11 00 FFCC 00 06 00 10 10 05 FFDA 00 08 01 01 01 00 00 3F 00 … FFCC 00 06 01 10 11 05 FFDA 00 08 01 02 11 00 3F 00 … FFDA 00 08 01 03 11 00 3F 00 … FFD9

17. Markers Code NameCategory FF01TEMFor temporary use in arithmetic coding FF02-FFBFRESReserved FFC0SOF 0 Nondifferential Huffman –coding frames Baseline DCT FFC1SOF 1 Nondifferential Huffman –coding frames Extended sequential DCT FFC2SOF 2 Nondifferential Huffman –coding frames Progressive DCT FFC3SOF 3 Nondifferential Huffman –coding frames Lossless sequential FFC4DHTDefine Huffman Table(s) FFC5SOF 5 Differential Huffman –coding frames differential sequential DCT FFC6SOF 6 Differential Huffman –coding frames differential progressive DCT FFC7SOF 7 Differential Huffman –coding framesdifferential lossless FFC8JPGReserved for JPEG extension FFC9SOF 9 Nondifferential arithmetic–coding framesExtended sequential DCT FFCASOF 10 Nondifferential arithmetic–coding framesProgressive DCT FFCBSOF 11 Nondifferential arithmetic–coding framesLossless sequential

18. Markers Cont.. FFCCDACDefine Arithmatic Coding FFCDSOF 13 Differential arithmetic- coding frames differential sequential FFCESOF 14 Differential arithmetic- coding frames differential progressive FFCFSOF 15 Differential arithmetic- coding framesdifferential lossless FFD0-FFD7RST m Restart with modulo 8 counter m FFD8SOIStart of image FFD9EOIEnd of Image FFDASOS Start of Scan FFDBDQTDefine quantization table FFDCDNLDefine number of lines FFDDDRIDefine restart interval FFDEDHPDefine Hierarchical progression FFDFEXPExpand reference images FFE0-FFEFAPP n Reserved for application use (two different applications may have conflicting usage of APP n ) FFF0-FFFDJPG n Reserved for JPEG extension FFFECOMComment

19. Frames and Scans Frame Header SOF n ParameterSymbol Size Marker ( FFC0-3, 5-7, 9-B,D-F)SOF n 16 Frame header lengthLf16 Sample precisionP8 Number of lines Y16 Number of samples per lineX16 Number of components in frame Nf8 Frame component specification Component identifierC i 8 Horizontal sampling factorH i 4 Vertical sampling factorV i 4 Quantization table destination selectorTq i 8 NfC1H1V1TqC2H2V2Tq2C3H3V3Tq3 030021 01110102 1101 The above example shows a three component system. YCrCb with identical sampling for all components in the vertical direction and twice as many Y samples as Cr or Cb samples in the horizontal direction. Let the first component use quantization table 0 and the rest use quantization table 1. If Y, Cr and Cb are numbered 0, 1, and 2 respectively, the components specification is as given above.

20. Other Header Structures Scan Header DHT Marker Segment DAC Marker Segment DQT Marker Segment Restart Segment

21. References JPEG still image data compression standard, William B Pennebaker, JL Mitchell, VNR, International Thomson Publishing Inc.