An application-oriented quality evaluation for SAR image

In this paper, we present a novel no-reference method to evaluate the quality of Synthetic Aperture Radar (SAR) images based on application. The features for predicting the perceived image quality are extracted by considering key human visual sensitivity factors such as mean, variance, contrast, and sharpness. In order to take into account the whole range of subjective quality from `excellent´ to `bad´, we construct retrogressive images by four means aiming to original, which are adding noise, reducing resolution, defocusing and multi-look to quantify the visual quality of SAR image. Image quality evaluation involves computation of functional relationship between human visual system (HVS) features and subjective assessment. Here, the functional relationship is approximated using back propagation (BP) network. The advantage of using BP network is its capability to learn new samples without affecting the past learning. Experimental results prove that the prediction of the trained BP network does emulate the subjective evaluation.