Study On The Human Vision Characteristics-Based Color-to-Gray Image Conversion Algorithms

With the continuous development of science and technology,more and more image processing algorithms and software continue to enrich and change our lives: from a variety of social media and software often used in graphic software to smart home,smart phones and other electrical equipment,cannot do without all kinds of image processing algorithms.And the color-to-gray image processing problem is a need for a number of practical applications in a basic image processing steps,the effect of good or bad on the subsequent treatment effect has a direct impact.In many areas also need to directly use the color-to-gray image algorithm,such as medical,agriculture,art,etc.,are directly used color-to-gray image processing algorithm.With the introduction of various theories and methods,the researchers have studied the color-to-gray image processing algorithm more and more thoroughly.The essence of color-to-gray image processing is the process of converting the three-dimensional color image data into the one-dimensional gray scale image data,and the information loss is inevitable during the conversion process.Therefore,in the process of color-to-gray image,how to preserve its visual features while preserving the information contained in the image is what we need to study and explore the problem.In order to further study the above problems,this paper attempts to extract the human visual features from two angles to deal with the color-to-gray image process——gray-scale domain and color-scale domain.The experimental results show that the Just Noticeable Difference(JND)of the gray-scale domain is better than the traditional fixed weight algorithm,but its stability and the original image information are not very good.The visual feature extraction algorithm of color field is more effective.Therefore,we improved two kinds of color-to-gray image algorithm which is ideal for the original image information——the principal component analysis(PCA)proposed by Seo et al and the maximized information entropy algorithm proposed by Wan et al.We take the feature extraction algorithm of color-scale domain to improvement them and achieved good results.In the subjective evaluation method commonly used in the algorithm,a more reasonable C2G_SSIM model is used to evaluate the objective qualitative evaluation.For the Wan et al algorithm,the color image gray scale algorithm based on color-scale domain visual feature extraction has good image effect compared with other algorithms in subjective evaluation and objective evaluation.