Color Image Data Compression Using Daubechies Wavelet Transform

Image processing requires an enormous storage, processing, and communication. To reduce this amount of data, image compression is necessary. We applied wavelet transform, which is an up coming technique and very powerful tools for signal processing, to color image compression. In this thesis, we use Daubechies 20 wavelet transform to the image compression procedure. There are five steps in the procedure. First, we extract color image into RGB components. Second, we transform RGB components to wavelet coefficients by using wavelet transform and pyramid algorithm. Third, we cut off some coefficients and scaled the results to eliminate less important data. Forth, we quantize wavelet coefficients by LBG vector quantization according to the significant of wavelet coefficients. Finally, we apply arithmetic coding, which is a lossless coding process, to wavelet coefficients. We test our algorithm with the images of Lena and Peppers. Lena is compressed with compression ratio 50-60 yielding PSNR around 34-35 and compression ratio 80-100 yielding PSNR around 30-31. Peppers is compressed with compression ratio 55-58 yielding PSNR around 34 and compression ratio 80-100 yielding PSNR around 30. Comparing with JPEG, this technique gives high PSNR than JPEG, at high compression ratio or compression ratio over 90, but less PSNR at low compression ratio or compression ratio below 85.