IMAGE COMPRESSION WITHOUT VISUALLY
VISIBLE DISTORTIONS
A.N. Zemlyachenko, V.V. Lukin
Methods for lossy compression of images without visual visible
distortions are considered. Metrics PSNR-HVS-M and MSSIM are used for
controlling visual quality of compressed images. The dependency of the compression
ratio on the values of the metrics PSNR-HVS-M and MSSIM preset by the user is
considered. The results of compression for different test images and coders are
presented. The comparative analysis is carried out, the
recommendations on the coder use are given. It is shown that the compression
ratio greatly depends on such features of the compressed images as complexity
of information content and level of the noise.
Keywords: compression ratio, lossy compression, image visual
quality metrics.