A novel approach to no-reference image quality assessment using canny magnitude based upon neural network

An immeasurable challenge in the field of image processing at the present time is image quality assessment. Image quality assessment techniques help in improving the visibility of an image by enhancing the brightness between objects and their backgrounds. Image quality assessments are on average performed as a contrast sketch followed by a pixel restoration. Various types of image quality assessment techniques are studied like full-reference(FR) image quality assessment, reduced-reference(RR) image quality assessment, no-reference(NR) image quality assessment with promising results, but no any one researcher finds the best and optimal solution. Many No-reference image quality assessment techniques have been proposed in existing work but our new proposed method through their feature learning approaches have outperformed. To improve the drawbacks of existing technique, we designed neural network framework that works on bigger portions of image. In this paper, we combine canny magnitude and feed forward back propagation neural network to obtain quality score based on image pixel scores and weights. The proposed method is used into no-reference image quality assessment as a classifier that can classify the set of best and appropriate pixels. However, the network structure of proposed method generates an effective model for estimating image quality. The proposed algorithm achieves good performance after experimental results on various image distortion types. The proposed method achieves excellent results on LIVE and TID2008 datasets.