Image compression

Abstract

The use of computers for handling images is dramatically growing, but digital images typically require a large number of bits to represent them. As a result, the problems of image storage, retrieval, and transmission have arisen. To overcome these problems, image compression techniques are needed to reduce images’ sizes. In addition, different applications have different requirements. Some applications, like medical imagery, require image compression without loss of image quality, while some applications, like World Wide Web imagery, require high compression ratio rather than high quality of images. This project aims to investigate a wide range of existing image compression algorithms, to design and to implement both lossless and lossy image compression schemes, in order to satisfy different kinds of applications. In this project, five lossless compression methods and four lossy compression methods are proposed. Then, their performance and compression ratio are evaluated and compared with each other. These compression methods are applying a wavelet transform to an image, and then removing some of the coefficient data from the transformed image by EZW, SPIHT and/or arithmetic coding. After testing a number of sample images, the best lossless image compression scheme can compress sample images up to about 53% of original size, while the best lossy image compression scheme can get around 15% of original size for Peak Signal-to-Noise Ratio (PSNR) values of 30 dB or more.