A Research Of Image Color Appearance Modeling Based On Cognition Processing Of Human Vision

The emergence of different types of imaging equipment makes the digital image original become more complex.The new challenge is put forward for the high fidelity color reproduction of cross media.Although the modern theories of color vision promoted the digital process of color science,a series of visual uniform color space and color appearance model have been produced.Those gradually met the needs of cross-media color production.But there are a lot of color visual phenomena that can't still get better explanation.But the digital original images are obtained from different imaging devices under different environmental conditions.The original image color appearance can't well be reproduced by means of the current ICC color management technology.That is to say,the visual color perception of the human visual system is difficult to reproduce.Therefore,the topic on the imaging mechanism of the human visual system is still a long-term research focus,which should be based on the simulation of human visual perception.This paper is devoted to the study of digital image cross media reproduction,to solve the high fidelity reproduction problems that the digital image come from the complex observation environment.So the imaging mechanism of human visual system will be researched,the goal is to establish the image color appearance model,development a visual color constancy computation algorithm,prediction the image color appearance under different environmental conditions.With development of visual physiology,anatomy,optic nerve electrical and so on,the color cognitive processes of human vision system has become more and more clear.The color perception not only has the color transfer multi-channel mechanism,but also there is the multi-channel mechanism of spatial frequency transfer.The nerve conduction of the retinal layer has been recognized that is not simply conducting the trichromatic signal to the opposition color signal,but there is the existence of the color antagonistic mechanism.Those neurons of visual pathway exhibit different receptive fields.The signal transfer between receptive fields is from lower to higher,and the advanced receptive field has reverse feedback to the lower.Based on the classical color vision theory and the results of the latest vision science,as well as the cross-media image reproduction,three-dimensional image acquisition,the high dynamic range image compression,the image quality metric and so on,the research that the image color appearance model simulating the human color vision is very meaningful.Therefore,the paper will establish an image color appearance model that related to the cognitive processes of human color vision,and develop the visual computing color constancy algorithm to predict the color constancy of human color vision.According to the modern opponent color theory,the paper suggests that the color constancy of human color vision shall be calculated in two stages,the one is light color adaptation,and the other is the color antagonistic contrast.The theoretical innovation of the dissertation has merged the color transfer multi-channel mechanism with the multi-frequency channel mechanism of spatial frequency transfer.The paper has built an image color appearance model that can predict the color constancy of human color vision.In order to study the multi-frequency channel processing mechanism of visual perception,the paper designed psychophysical experiments,measured the spatial frequency tuning curves of the opponent color black-white,Red-Green and yellow-blue.Based on the Retinex theory,those spatial frequency tuning-curves were designed for several filter operators of different spatial dimension.The multi-scale Retinex theory has been developed.The center-surround contrast mechanisms of the human color perception is simulated.The spatial frequency multi-channel process is dividing the image into several frequency channels image according to the spatial frequency characteristic of image.This described the image comparative analysis of global and local.The color transfer multi-channel process is the optic nerve signal transmission from the tricolor photosensitive signal to the opposition color mechanism.In trichromatic phases the exposure was adjusted,and in opponent color phase the spatial contrast was predicted.The sensitivity of the spatial complex color stimuli related to the spatial resolution.The multi-channel of color transfer and multi-channel of spatial frequency transfer are merged that the image color appearance model can be achieved availability.The application innovations of the dissertation are mainly in three aspects: First,the spatial frequency transfer characteristic of the human color vision has been measured,which is on the lightness axis and in the chromaticity axis plane of the opponent color space CIELAB.Those contrast sensitivity function curves have been fitted including five brightness values and twelve saturated colors.Those different contrast sensitivity functions were used to describe the color discrimination of human vision in CIELAB color space.Second,based on the spatial frequency tuning-curves,the ten low-pass filter operators have been designed,which associate with the viewing condition.Those low-pass operators use to simulate the multi-scale features of the color signal processing of opponent color channel.The regulation of the operator is not only in the spatial dimension,but also relates to the center frequency in the range of adjustment.The contrast between the center pixel and its' surrounding pixels was get the weighted adjustment.Third,based on the theoretical innovation model,the color constancy algorithm has been proposed.The algorithm simulated the signal process of human color vision to convert the trichromatic to the opposite color.The light color adaptation was calculated in the exposing stage of the three-primary colours,the white-image of the adaptation field was estimated.In processing stage of the opponent color,the visual antagonistic mechanism of "Center-Surround" was simulated to complete the calculation of color contrast.The experimental results show that the color constancy algorithm related to human visual cognitive process can predict the image color appearance reproduction under a variety of human visual scene.The performance of the algorithm is to meet the needs of high fidelity printing of cross-media image reproduction.The algorithm can also be applied to the industry related to the visual image imaging technology,such as 3D image acquisition,high dynamic range image compression,image quality metric and so on.It is very meaningful to simulate the image color appearance model of human visual mechanism,and the application prospect is very extensive.