MOS prediction of SPIHT medical images using objective quality parameters

Abstract

Correlating objective and subjective quality assessment parameters of compressed digital medical images has been an open challenging problem in tele-radiology. Establishing this correlation is crucial in determining the upper limit of image compression threshold for preserving diagnostically relevant information based on mean opinion score (MOS). This paper presents a suitable method for finding correlation between PSNR and Structural Similarity (SSIM) index objective image quality parameters with subjective MOS for SPIHT [4] compressed medical images based on six independent observers. The suggested method can be potentially used for deciding upper compression thresholds for medical images. It is found that correlation coefficient (CC) between the PSNR and MOS for CT scan and MRI images are 0.979 and 0.960 respectively whereas their corresponding values are 0.868 and 0.955 considering SSIM. Further, MOS prediction models have been proposed considering PSNR and SSIM which closely match with the subjective MOS. © 2009 IEEE.