1. Wavelet Based Color Compression Greg Beranek

2. Background • WDR • Significance Pass • With threshold T and T2 find item where T2>Item>=T • Encode the Address • Encode Plus or Minus • Repeat • Refinement Pass • If Item >=T2 • Write Refinement • Repeat

3. WDR

4. Color WDR • Significance Pass • Three Passes • Red Pass • Green Pass • Blue Pass • Refinement • Three Passes • Red • Green • Blue

5. Color WDR

6. Issues Color WDR • Progressive nature • Cannot stop anywhere (missing color component) • Speed • 3 Passes • 3 Scan orders • 3 data matrices • Accounting

7. Resolution To Color WDR Issues • Color Correlation (lack of correlation) • Ability to predict significant values • Encode more data with one color and get a very close color when decoding • Simplify scan order

8. Color Space • Primitive Color Spaces • RGB • Additive color space • Tells how much light is needed to make a specific color • CYM K • Subtractive color space • Tell what quantity of pigment is needed to make a specific color • Other Color Spaces • YUV • YPbPr (YCbCr) • Video and Image Compression Schemes, Analog (Digital) • YCoCg • Integer to Integer Color space similar to YPbPr

9. Transform Matrix RGB to YCoCg • Forward - RGB -> YCoCg • Reverse - YCoCg -> RGB

10. Back To Color WDR • We now have the ability to: • Predict significant color values • Simplify to one scan order • Retain some RGB color information when we stop encoding anywhere

11. Color WDR

12. Optimizations • Encoder optimizations • ASWDR • Lossy Color Space

13. How good is this compared to JPEG 2000?

14. Pictures Lena Left : Original (Cropped) Top Right : Color WDR 1 bpp Bottom Right : JPEG 2000 1 bpp

15. Pictures Hats Left : Original (Cropped) Top Right : Color WDR 1 bpp Bottom Right : JPEG 2000 1 bpp

16. Pictures Parrots Left : Original (Cropped) Top Right : Color WDR 1 bpp Bottom Right : JPEG 2000 1 bpp

17. Questions?