On the Efficiency of the Wavelet Based Compression Techniques

In this paper we address the problem of the correlation between the objective quality measures, like mean squared error or signal to noise ratio and the subjective quality perceived by human viewers. Using both the objective and subjective quality measures we aimed to compare the efficiency of the wavelet based lossy codecs DJVU and LWF with the DCT based standard JPEG compression method. For the experiments a set of 3 gray-scale and 3 color test images was used. The images were encoded at the same compression levels and standard objective quality measures were calculated for all images. For the visual assessment of the quality of the compressed images, a group of 100 male students evaluated them using a linear rating scale. The results have shown that there is no significant correlation between the objective and subjective quality measures. Although the wavelet based LWF codec outperformed significantly DJVU in terms of MSE and SNR, no significant differences in human visual perception of those images were observed. It was also observed that the histogram of the viewers preferences regarding the image compression ratio is bimodal and no optimal compression level could be established.