Color Matching Algorithm And Application Research Integrating Visual Perception

As the first element affecting human vision,color is particularly important in the process of product design and art creation.Reasonable color matching relationship can often enhance the added value of the product.In practical design processes,the efficiency of manual color matching often fails to meet the cycle requirements of current product design.In recent years,thanks to the rapid development of digital and information technology,color matching methods based on deep learning technology have become the focus of current research because of their ability to improve the efficiency of color matching.However,color matching is also constrained by physiological and psychological factors while following physical laws.How to establish a color matching model based on deep learning technology which is more consistent with visual perception has become the focus and difficulty of current research.In this thesis,we conduct an in-depth study on the following.(1)To address the issue of traditional dominant color extraction algorithms being unable to adaptively obtain the dominant colors in images.This thesis builds an adaptive dominant color extraction model for images based on the silhouette coefficient method.In the model,the silhouette coefficient values of different dominant colors are calculated during the pixel clustering process,and adaptive dominant color extraction is achieved by selecting the color with the maximum silhouette coefficient value.Experimental results show that compared with the classical K-Means algorithm,the median cut method,and the octree algorithm,the proposed model not only achieves adaptive extraction of dominant colors but also significantly improves the quality of dominant color extraction.The peak signal-to-noise ratio(PSNR)and structural similarity(SSIM)index of the images processed by the proposed model have been improved by 1.127 d B,4.217 d B,3.345 d B,and 0.027,0.041,0.039,respectively,compared to the classical algorithms mentioned above.Additionally,the subjective preference level indicators were improved by 4.22%,19.50%,and 22.96%.(2)To address the issue of the lack of effective visual perception ability in color matching evaluation based on a single feature similarity.This thesis proposes a color matching evaluation method that combines visual perception and similarity measurement.This method first establishes a minimum color difference model for evaluating the similarity of color palettes and performs feature-level fusion with the structural similarity of the corresponding images.Finally,this thesis uses eye-tracking technology to obtain visual perceptual measures of paired images and optimizes the weighting coefficients of feature fusion based on these measures.Experimental results show that the constructed color matching evaluation index has an average correlation coefficient of 0.808 with visual perception measures,which can provide support for quantitative analysis of color matching algorithms.(3)To address the issue that the color matching algorithm based on Pix2 Pix image translation model lacks visual perception data drive.This thesis proposes an intelligent color matching algorithm that incorporates visual aesthetics.The algorithm first uses eye-tracking technology to obtain visual aesthetics data flow.Then,it uses the data flow to drive the Pix2 Pix image translation model and uses the above color matching evaluation method for quantitative analysis.Experimental results show that the quantitative analysis index and the peak signal-tonoise ratio of the image are improved by 0.096 and 6.12 d B,respectively,compared with the Pix2 Pix image translation model,and the overall preference of the subjective evaluation is improved by 9.28%,which is closer to the Ground Truth Palettes matching effect.This thesis systematically analyzes the physiological and psychological attributes of color matching schemes,and proposes a color matching evaluation method that integrates visual perception and similarity metrics for the first time.At the same time,the Pix2 Pix image translation model driven by visual perception data flow is used to realize color intelligent matching.This thesis verifies the validity and advancement of the multi-dimensional color matching evaluation index constructed.The research of this paper has guiding significance for color matching in product design and artistic creation,which can adapt to the market demand of "small batch,multi-variety,fast delivery time",and has important theoretical research and application value.