Abstract:

As image structure feature is closely related to the human visual system, a new method based on structure feature was proposed to improve the result consistency with a subjective visual effect. First, image structure features were extracted though experience. And then various structural components were calculated with different weights, according to different sensitivities of eyes to image regions. Finally, a new model was proposed with linear regression. Experimental results on different kinds of images showed that the new method had high efficiency and accuracy compared to the PSNR and RMSE, and had a better consistency with a subjective visual effect.

Key words: image quality assessment, linear regression, human visual system, structure feature