Image Enhancement Method Based On Visual Perception

With the development of image information technology,color imaging devices and display are more and more widely used.Psychophysical experimental results show that the human eyes are more inclined to the image with high contrast and color.Image enhancement technology can enhance the image contrast,rich image color and improve the image visual effect,has become the core competitiveness of the display industry.The core difficulty of image enhancement technology is that it may introduce artifact,such as,over-enhancement,halo,unnatural,noise,etc.,it may leads to the decrease of image visual quality.In order to obtain the optimal perceptual enhancement image and avoid the significant artifacts,this paper focus on how to enhance the image based on the human visual perception.In this paper,a multiscale image enhancement method based on human visual preference is proposed to overcome the over-enhancement and halo effect.In this paper,we use statistical and psychophysical methods to estimate the perceptual preference.Firstly,a series of subjective psychophys ical experiments were used to collect the subjective preference data;then,using the Thurstones' law to divided statistical data to indicate the preference level of human eyes to different frequencies;finally,a multi-scale image enhancement model based o n perception is proposed to guide the multi-scale image enhancement.In this paper,we study the preference characteristics of human eyes to different frequency details,the results of the human eye preference are used to guide the image enhancement,and balance the relationship between the contrast and the enhancement of the image.As a nonlinear system,the visual system determines the different image content needs to adopt different enhancement methods to obtain the optimization of the perceptual quality,the content of the image varies a lot,this paper mainly focuses on the noise information,the low il umination area,skin color area.(1)For the noise in the image,in order to enhance the robustness of the algorithm,the non-local mean filter is used to preprocess before the image is enhanced.(2)For the low contrast area in the image,the traditional SSR and MSR can enhance thedetail of the image by removing the illumination layer,weaken the image depth,light direction confusion,lose the naturalness of the image,for addressing this shortcoming,a new image decomposition method is proposed,and preserve the illumination layer information.In addition,the brightness of the illumination layer is adjusted by using the logarithm histogram.To enhance the contrast of low luminance region under the condition,keeping the original luminance distribution,and restrain the highlight area contrast overflow.(3)In order to get better color image visual effects,considering that the human eye has a large amount of prior knowledge and a high degree of sensitivity,a color saturation stretching algorithm based on skin color protection is proposed,the algorithm can protect the human eye highly sensitive skin color of the region to achieve a better visual effect.Finally,this paper analyzes the effect of the algorithm from the subjective and objective evaluation,and proves that the algorithm has good visual effect and naturalness.