Image restoration based on visual contrast sensitivity and just noticeable distortion perception
Yang Ya-we · Optics and Precision Engineering · 2014
As traditional restoration algorithms have weaker perceptual opinion in a low Peak Signal Noise Ratio(PSNR),a novel image restoration approach based on visual contrast sensitivity and Just Noticeable Distortion(JND)perception was proposed.Firstly,the image was coded as a sets of bands in different scales with Discrete Cosine Transform(DCT),the JND threshold of each band was calculated with probability summation and DCT perception model,and distortion measure was calculated with a Minkowski metric.Then,an iterative bisection procedure was used to adjust the channel weight for each band in each scale until the perceptual distortion of band was just equal to the antici-pant distortion threshold.Finally,the coefficient of each band was weighted with visual contrast sensitivity function,and the coefficient after visual perceptual weighting was reconstructed through an inverse DCT to gain the restoration image.Experimental results based on LIVE image database show that this algorithm can effectively enhance PSNR and improve perceptual quality of the recovery images.