Saturation quality assessment model of stereoscopic image base on Gaussian white noise

In the stereoscopic image quality assessment, the overwhelming majority of previous studies convert color images to gray scale images, which loses the color information in order to reduce the complexity. However, the loss of color information is not conducive for color stereoscopic images to make the right assessment. To solve this problem, saturation quality assessment model of stereoscopic image base on Gaussian white noise was proposed. Firstly, extract saturation gradient values of eight directions around each pixel from reference images and distorted images. Secondly, euclidean distance of eight gradient feature vectors are calculated between gradient values of reference image and gradient values of distorted image. Euclidean distance are taken as image quality evaluation indexes. Thirdly, using curve model for fitting the mapping relationship that between indexes and DMOS. Finally, eight directions of indexes are used to multiple linear regression analysis and the regression equation is saturation quality assessment model of stereoscopic image. The method was tested on the LIVE 3D Image Quality Database published by university of Texas. The Linear Correlation Coefficient (LCC) and Spearman Rank Order Correlation Coefficient (SROCC) achieved 0.917 and 0.918.The results have high accordance with the subjective evaluation.