P~2SNR: Perceptual Full-Reference Image Quality Assessment for JPEG2000

摘要

Estimation of image quality is decisive in the image compression field. This is important in order minimize, induced error via rate allocation[l]. Traditional full-reference algorithms of image quality try to model how Human Visual System detects visual differences and extracts both information and structure of the image. In this work we I propose a quality assessment, which weights the mainstream PSNR by means of a perceptual model (P~2SNR). Perceptual image quality is obtained by estimating the rate of energy loss when an image is observed at monotonically increasing distances. Experimental results show that 'P2SNR is the best-performing algorithm, compared with another eight metrics such as MSSIM, SSIM or VIE, among others, when an image is distorted by a wavelet compression. It has been tested across TID2008 image database.