Reduction of blocking artifacts in DCT-coded images

1. Reduction of blocking artifacts in DCT-coded images Presenters : 695410076 梁速清 695410122 莊令誼 Date : 2007.6.20

2. Outline • Introduction • Luo’s Method [1] • Shi’s Method [2] • Comparison • Conclusion 1

3. Introduction DCT & Quantize IDCT Original Lena Lena after DCT 2

4. Luo’s Method (1/4) 一. Search two adjacent blocks(8x8),Block A and Block B. 二. Assume Block A and Block B have the same pixel value, a, b respectively ,and a ≠ b. 三. Block C include right half of Block A and left half of Block B 3

5. Luo’s Method (2/4) * DCT coefficient of Block A , Block B, Block C Block A Block B Block C 4

6. Luo’s Method (3/4) * Ying proposed three constraint : (a, b exist strong edge. c means low frequency) * T1 = 350, T2 = 120 ,T3 = 60 . * The choice of thresholds are image and compression ratio dependent. 5

7. Luo’s Method (4/4) * The blockiness reduction in the DCT domain by modifying the five relevant coefficients of Fc(u,v). * Block C is modified by the weight average of block A,B, and C. * The pixels of C(k,l), are replace by CMc(k,l), the inverse DCT of FMc(u,v) 6

8. Shi’s Method • Propose a visibility function of blocking artifacts based on HVS (Human Visual System ) • Reduction of Blocking artifacts • Smooth regions : modify parameter, use linear blocks 7

9. Model for Blocking Artifacts • New shifted block c • Step function 8

10. Model for Blocking Artifacts (cont.) • Model of c : average value of block c : the magnitude of s : the activity inside c • C, S represent the DCTs of c, s 9

11. Visible Function • Visible function falls below a threshold, no further process need • Mean squared difference of slopes (MSDS) c( i , j ) : luminance value of pixel ( i , j ) 10

12. Visible Function (cont.) • Activity Masking Ah : horizontal activity Av : vertical activity • Brightness Masking b0=150, r =2 b : average luminance value 11

13. Reduction of Blocking Artifacts • Edges should be detected to avoid blurring information-bearing edges in the image • Edge detection : satisfy the two conditions, represent no edges QP : quantization parameter 12

14. Reduction of Blocking Artifacts (cont.) 13

15. Reduction of Blocking Artifacts (cont.) • For smooth region • Replace step block s with linear block in theshifted block 14

16. Reduction of Blocking Artifacts (cont.) • Use information of the left block a and the right block b 15

17. Reduction of Blocking Artifacts (cont.) • A DCT-domain post-filtering method is applied to eliminate newly-created artifact • For edge region • The same post-filtering method is applied 16

18. Shi’s Result 17

19. Zero Liu Shi Shi’s Result (cont.) • Bitrate = 0.183 18

20. Comparison (Luo’s & Shi’s) Luo Shi 19

21. Conclusion 20

22. Reference • [1]Ying Luo and Rabab Kward, "Removing the blocking artifacts of block-based DCT compressed images," IEEE Trans on Image and Processing, vol. 12, pp.838-842, July, 2003 • [2]M. Shi, Q. Yi, and J. Gong, “Blocking Effect Reduction Based On Human Visual System For Highly Compressed Images,”Electrical and Computer Engineering, Canadian Conference on, May 2006 • [3]S. Z. Liu and C. B. Alan, "Efficient DCT-domain blind measurement and reduction of blocking artifacts," IEEE Transactions on Circuits and Systems for Video Technology, vol. 12, pp.1139 - 1149, Dec., 2002 21