LGPS: Phase Based Image Quality Assessment Metric

Phase map of the images captures the most fundamental cognitive features and thus is widely used in various digital image processing tasks. In this paper, we propose the Log Gabor Phase Similarity (LGPS), a novel full reference image quality assessment metrics based on measuring of similarities between phases in log Gabor transform domain. Phase can capture any changes in image details regardless of the fluctuation in contrast, and the similarity between phase maps provides a measure of the perceptual quality of images. An image is firstly decomposed by a filter bank consisting of a pair of log Gabor filters. The phase maps are then computed from the responses of each filter pair. We have developed a window-based similarity metric to evaluate the resemblance between phase maps so as to measure the quality of the image. Experimental results and comparative studies suggest that LGPS can be used to predict the perceived quality of images with different distortions.