Image compression with wavelet transforms and set partitioning in hierarchical trees method

Abstract

Digital images are claiming an increasingly larger portion of the information world. In this study, wavelet transform is chosen for image compression. Image is decomposed into subbands of averages and details with wavelet transform. Obtained detail subbands have small valued coefficents that also constitute a small percentage of image energy. If these small valued coefficents are quantized to zero with a chosen threshold, there will be no great loss in the image. This feature provided by wavelet transform is a basis for set partitioning in hierarchical trees algorithm. With this algorithm discrete wavelet transform coefficents are organized in spatial orientation trees. The reason to have formed these trees is gathering all pixels that are highly correlated with each other. With this kind of set structure, the similarity among the coefficents in a set from one level to the next is increased. The main aim in this algorithm is to code the trees with highly correlated, zero valued coefficients with a single code word. With these features in the nature of wavelet transform and set partitioning in hierarchical trees algorithm, it is intended to get high compression ratios in images.