NO REFERENCE IMAGE QUALITY ASSESSMENT BASED ON LOCAL BINARY PATTERN STATISTICS

Multimedia, including audio, image and video, etc, is a ubiquitous part of modern life. Evaluations, both objective and subjective, are of fundamental importance for numerous multimedia applications. In this paper, based on statistics of local binary pattern (LBP), we propose a novel and efficient quality similarity index for no reference (NR) image quality assessment (IQA). First, with the Laplacian of Gaussian (LOG) filters, the image is decomposed into multi-scale sub-band images. Then, for these sub-band images across different scales, LBP maps are encoded and the LBP histograms are formed as the quality assessment concerning feature. Finally, by support vector regression (SVR), the extracted features are mapped to the image's subjective quality score for NR IQA. The experimental results on LIVE IQA database show that the proposed method is strongly related to subjective quality evaluations and competitive to most of the state-of-the-art NR IQA methods.