Color Image Indexing Using BTC Author: Guoping Qiu Source: IEEE - PowerPoint PPT Presentation

Color Image Indexing Using BTC Author: Guoping Qiu Source: IEEE Transaction on Image Processing, Vol. 12, NO. 1, pp. 93-101, 2003 Speaker: Tzu-Chuen Lu

Outline � Color Image Retrieval � Color Correlogram (CC) � Block Truncation Coding (BTC) � BTC Color Co-Occurrence Matrix (BCCM) � Block Pattern Histogram (BPH) � Experiments and Results � Conclusions

Color I mages Retrieval – Color Histogram 5 0 3 0 2 5 4 0 2 0 3 0 1 5 2 0 1 0 1 0 5 0 0 1 0 3 5 6 5 1 5 0 2 0 0 1 0 3 5 6 5 1 5 0 2 0 0

Color Correlogram (CC) Image: I ’ Image: I 150 150 65 10 10 155 142 50 10 12 20 22 70 75 100 10 35 65 65 150 31 175 80 34 32 35 150 65 35 35 221 5 88 45 31 200 10 65 35 35 230 12 30 20 88 200 10 35 10 65 1 0 3 5 1 5 0 2 0 0 1 0 3 5 1 5 0 2 0 0 … … 1 2 3 4 1 2 3 4 … 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 … 1 2 3 4 1 2 3 4 Dis. Dis. 1 0 3 2 4 3 1 0 … … 3 5 0 1 2 3 … 3 5 … 6 5 6 5 … … 1 1 5 5 0 0 3 2 6 1 1 … … 1 1 2 2 0 0 0 0 5 1 0 2 3 … …

Block Truncation Coding (BTC) Image: I Average B1 B2 1 2 6 1 0 3 1 5 1 4 1 . 5 B3 B4 Bitmap: I ’

BTC for Color Image Coding Bitmap: I ’ Image: I 8* 16 = 128 (bits) 16 + (4* 2)* 8 = 80 (bits) Mean Values

BTC Color Co-Occurrence Matrix (BCCM) 0: 55 0: 24 1: 234 1: 228 0: 91 0: 18 1: 210.5 1: 112 10 1 20 2 … … 100 228 … … 255 200 3 2 3 5 10 1 … … 1 0 1 1 20 2 … … … … … … … … … … 2 1 3 3 200 24 … … 1 … 255

For Example 10 20 30 40 � I 1 5 3 0 0 10 0 2 1 0 20 2 1 0 6 0 30 0 2 0 1 40 � I 2 10 20 30 40 2 0 0 3 10 1 0 1 1 1 20 6 8 1 1 30 0 2 3 0 40 − − − | 5 2 | | 3 0 | | 1 0 | = + + + Dis tan ce ... + + + + + + 1 5 2 1 3 0 1 1 0

Block Pattern Histogram (BPH) Bitmap: I ’ 1 0 1 1 1 1 1 1 1 0 1 1 0 0 1 1 1 0 1 1 0 1 0 1 0 0 0 0 1 1 0 1 1 0 1 1 1 1 1 1 1 0 1 1 0 0 1 1 1 0 1 1 0 1 0 1 0 0 0 0 1 1 0 1 1 2 3 … 256 1 2 5 6 … Indx. Indx. 256 5 2 1 2 3 0 Pr. Pr. 1

For Example 10 20 30 40 � I 1 5 3 0 0 1 2 3 … 256 10 Indx. 0 2 1 0 20 5 2 1 2 3 0 Pr. 2 1 0 6 0 30 0 2 0 1 40 � I 2 10 20 30 40 2 0 0 3 1 2 3 … 256 10 Indx. 1 0 1 1 1 20 3 1 0 3 Pr. 6 8 1 1 30 0 2 3 0 40 − − − | 5 2 | | 3 0 | | 1 0 | = + + + + Dis tan ce ( ... ) + + + + + + 1 5 2 1 3 0 1 1 0 − − − | 5 3 | | 2 1 | | 30 0 | + + + ( ... ) + + + + + + 1 5 3 1 2 1 1 30 0

Experiments and Results - I • 720 texture images • 120 texture classes • 300* 200 pixels • A block with 100* 100 pixels

Experiments and Results - I

Experiments and Results - II 20,000 color images in DB Query Data

Experiments and Results - II 96 query images

Conclusions � Using a well known image coding technique BTC to retrieve images � Two image features derived directly from an image � The new method achieves coding and retrieval simultaneously