No-reference quality metric for contrast-distorted image based on gradient domain and HSV space

Abstract

Image quality assessment (IQA) plays an important role in digital image forensics. Due to the occurrence of contrast distortion during image acquisition and manipulation, IQA for contrast is a major issue. And it is vital for benchmarking and optimizing the image tampering detection and contrast-enhancement algorithms. In this paper, a new no-reference/blind image quality assessment (IQA) metric is proposed for evaluating image contrast. This research seeks for the inter-relationship between contrast distortion and visual perception quality. The comprehensive quality metric is obtained by combining local binary pattern (LBP) descriptor on gradient domain with color moment on HSV color space. And a prediction model is trained with support vector regression (SVR). Extensive analysis and cross validation are performed on four contrast relevant image databases, which validates the superiority of our proposed blind technique over state-of-the-art no-reference IQA methods.