No-reference image SNR assessment under visual masking

A novel no reference image signal to noise ratio(SNR) assessment method based on the property of human visual system (HVS) was proposed to evaluate the quality of noisy image. Firstly, the method adopted Hosaka plots idea and quadtree decomposition which wiped out the limits on image size. Secondly, this improved method was applied to the image to make plots and was employed to abstract sub-blocks noises. Then, the no reference image signal to noise ratio (NRPSNR) was computed according to the pollution degree. The method was tested and verified by using the LIVE Database Release2 image database. The experiment results show that this no reference assessment method which CC reaches -0.895 0, SROCC reaches -0.923 6, OR reaches 0.042 9，is superior to other three kinds of methods on the whole, it is more consistent with subjective image quality assessment, suitable for other noisy styles and embodies the HVS visual masking effect. It can give accurate and reliable quality assessment of no reference noisy images.