Anzhou Hu, Rong Zhang, Xin Zhan, and Yin Dong
Image quality assessment, human visual system, image content partition
The development of reliable objective image quality assessment (IQA) metrics coordinate to the human‟s perception is crucial in numerous image processing applications. State-of-art perceptual IQA methods focus on two techniques using the sensitivities of human visual system (HVS). One is perceptual pooling strategy in spatial domain while the other is multi-channel model in spectral domain. In this paper, we incorporate the two directions and propose a novel method which employs the interaction of spatial and frequency sensitivities of HVS. Experimental results demonstrate that the proposed metric achieves a better subjective perception consistent than those using Structural Similarity (SSIM), Multi-scale SSIM (MS-SSIM) and singular value decomposition (SVD). The metric also outperforms the methods only considering the spatial or spectral sensitivities of HVS.
Important Links:
Go Back
