Metric of fine structures distortions of compressed images

The article proposes a new Metric of Fine Structures Distortion (MFSD) in compressed images. The peculiarity of the metric is that it is based on the algorithm for identifying fine image structures using a normalized N-CIELAB system, which allows us to estimate distortions while taking into account the properties of contrast sensitivity of the eye. Experimental results of the estimation of distortions of test images are presented depending on the degree of compression and the quality parameter in the JPEG and JPEG2000 codecs according to the MFSD metric and the traditional metrics PSNR and SSIM. Based on the results of a comparative analysis of MFSD with subjective assessments of the quality of the compressed JPEG and JPEG2000 images, a new objective criterion for the high quality reproduction of fine structures of compressed images was obtained. The theoretical and experimental justification of the objectivity of the new criterion is presented based on the results of processing and analysis of distortions of the compressed test photorealistic images.