1. IMAGE COMPRESSION USING BTC Presented By: Akash Agrawal Guided By: Prof.R.Welekar

2. Contents Image compression Block Truncation Coding(BTC) Absolute Moment Block Truncation Coding (AMBTC) Minimum Mean Square Error (MMSE) Adaptive Block Truncation Coding (ABTC) Improved Adaptive Block Truncation Coding (IABTC) Results And Discussion Future Scope Conclusion References

3. Image compression What is the image compression ? To store the image into bit-stream as compact as possible and to display the decoded image in the monitor as exact as possible.

4. Compression can be categorised in two broad ways: Lossless Compression: Zero length suppression Run length encoding Pattern substitution Lossy Compression: Transform coding Vector quantization BTC

5. Block Truncation Coding(BTC)

6. In BTC, two statistical moments a and b are computed using the equations (3) and (4). BTC can be explained with following example.

7. The difference between the original image and reconstructed image is called Mean Square Error (MSE) and is calculated using the equation (5). The quality of the reconstructed image called the Peak Signal to Noise Ratio (PSNR) is calculated using the equation (6)

8. Absolute Moment Block Truncation Coding (AMBTC) In this technique image compression is done using absolute moment block truncation coding. It is an improved version of BTC, preserves absolute moments rather than standard moments. The statistical moments a (lower mean) and b (higher mean) calculated using the equations, (7) and (8). The coding and decoding processes are very fast for AMBTC because the square root calculation and multiplications are omitted.

9. Minimum Mean Square Error (MMSE) MMSE is the iterative process of AMBTC. This technique is used to reduce MSE value. In this method, the threshold value t is initialized by the average of minimum and maximum values of each block. The threshold value thus calculated is optimized through iterations as described in the Figure. The optimization process is terminated when the threshold values of consecutive iterations converge.

10. MMSE method

11. Adaptive Block Truncation Coding (ABTC) Adaptive Block Truncation Coding (ABTC) is based on multilevel quantizer. The input blocks are classified into three groups depending on the inter-pixel correlation within each block. (1) Low activity blocks (2) Medium activity blocks (3)High activity blocks Three set of quantizers are used in the ABTC method for all blocks.

12. a 4-level MMSE quantizer for high activity edge blocks. Let the output levels of the quantizers be a, b, c and d and are calculated as Xmin, (2Xmin+Xmax)/3, (Xmin+2Xmax)/3, and Xmax. Use the same iterative processes to optimize the thresholds t1, t2 and t3 for the output levels a, b, c and d and vice versa. The results of the quantization will be a bit plane whose individual elements are of size 2 bits and the output levels are a, b, c and d.

13. Improved Adaptive Block Truncation Coding (IABTC) The proposed method (IABTC) is based on Adaptive Block Truncation Coding and is used in further reducing the bit rate and to improve image quality. All the input image blocks are categorized into three groups based on sum value (S), which is calculated using the equation (9). The blocks are categorized in to three groups as follows:

14. t1,t2 and t3 is calculated as a,b, c and d is calculated as

15. The compression of image is done in two levels in IABTC: I. First level compression Step1: Input the image size n x n pixels Step2: Divide the image into N blocks, each of size 4 x 4 pixels. Step3: Compute the mean for each block using the equation (1). Step4: Compute S for each block using the equation (9) Step5: Categorize the block (Low or Medium or High Detailed block) using the equations (10). Step6: Store the set {B, statistical moments} II. Second level compression

16. Bit-rate calculation: The bit plane is generated by coding the individual pixel as the bpp of the image compressed using IABTC method is calculated as o n=n1*6+n2*(16+6+6) + n3*(32+6+6) o bpp = n / (256*256)

17. Results And Discussion Experiments were carried out with standard images Cameraman,Boats, Bridge, Baboon, Lena and Kush of size 256 x 256 pixels. (a) Cameraman (b) Boats (c) Baboon The algorithm is implemented using MATLAB 7.0 on Windows Operating System. The hardware used is the Intel® 3.06GHz Processor with 512 MB RAM.

18. BTC AMBTC ABTC PSNR: 34.33 BPP=2.00 PSNR: 32.13 BPP=2.00 PSNR: 32.07 BPP=2.00 MMSE IABTC PSNR:32.70 BPP=2.00 PSNR: 34.67 BPP=1.01 Reconstructed images

19. The PSNR values and bit-rate obtained with different BTC methods and IABTC ImagesBTCAMBTCMMSEABTCIABTC BPPPSNRBPPPSNRBPPPSNRBPPPSNRBPPPSNR Cameraman 232.07232.13232.701.9030.191.013467 Baboon 237.14237.46237.682.1337.570.9736.46 Boats 232.77233.15233.811.8230.361.2434.74 Bridge 230.68230.94231.472.0931.001.7432.02 Lena 234.62234.85235.392.1835.600.9835.79 Kush 235.41235.55236.111.8834.501.0035.77 Average 235.78234.01234.532.0033.201.1634.91

20. Future Scope The proposed technique can also be applied to color images. This method is suitable for implementation in small hand held devices where image processing service has become inevitable. The bpp rate can further be reduced. The Psnr value can be increased.

21. Conclusion In this paper, the feature of inter-pixel redundancy is exploited to reduce the bpp to significant level. The average bpp obtained with the proposed method is 1.16.

22. References S.Vimala, P.Uma, B.Abidha “Improved Adaptive Block Truncation Coding for Image Compression”, International Journal of Computer Applications (0975 – 8887) Volume 19– No.7, April 2011. Meftah M.Almrabet, Amer R. Zerek, Allaoua Chaoui an Ali A. Akash,” Image Compression using Block Truncation Coding”, IJ-STA, Vol 3, No.2, PP.1046-1053, December 2009. Maximo D.Lema, O.Robert Mitchell, ”Absolute Moment Block Truncation Coding and its Application to Color Images”, IEEE Transactions on Communications, VOL. COM-32, NO. 10, October 1984. Lucas Hui, “An Adaptive Block Truncation Coding Algorithm for Image Compression”, IEEE, PP.2233-2235, 1990

23. Thank You..