Comparative study of lossy image compression techniques

Abstract EN

This paper presents a comparative study of both spatial domain and transform based loss image compression techniques.

Spatial domain methods include improved gray-level scaling, loss predictive coding, block truncation coding, and a new quantization technique based on the rate distortion function which is introduced here.

Transform based coding techniques include discrete cosine transform with maximum variance selection, JPEG compression, and wavelet based compression.

These techniques are reviewed and their performances are compared with respect to average compression ratio, the average compression time, average relative perceptual distortion and compression consistency over nine test images of different sizes and different levels of detail.