Research On Color Transfer Method Based On Local Feature Matching

Image enhancement technology is a key technology in the field of image processing,which has penetrated into all aspects of people's lives.Among them,color transfer as an important means of image enhancement has been focused by researchers in recent years.The color transfer technology can transfer the color characteristics of the reference image to the content image,so that the resulting image has the structural characteristics of the content image and the color characteristics of the reference image,which can better improve the artistic style and visual effect of the image.Traditional color transfer methods often have problems such as blurring of details,enhanced noise,lack of sense of hierarchy,and inability to maintain the color characteristics of the reference image well.In addition,most of the existing objective evaluation indicators can't evaluate the effect of color transfer well.To this end,the article mainly did the following work:(1)Aiming at the problems of lack of hierarchy,insufficient feature retention and poor visual effects in traditional color transfer methods,this paper proposes a color transfer method based on local feature matching based on image signature and optimal transmission.Firstly,the method of salient region detection based on image signature is introduced to obtain the foreground and background regions of the reference image and content image,and then feature matching is performed in the corresponding regions to improve the layering of the result image while transferring colors.Secondly,based on the traditional linear color transfer model,combined with the optimal transfer theory,an improved color transfer strategy is proposed to further improve the quality of the resulting image.A larger number of experimental results show that compared with the traditional color transfer method,the result image obtained by the method proposed in this article can better maintain the edge structure,layering sense of the content image and the color characteristics of the reference image,and has a better visual effect.(2)Aiming at the problems of unreasonable region segmentation,insufficient layering,and inability to maintain the content image structure in the color transfer method based on color clustering,a new color transfer method based on random walk image segmentation is proposed.First,an improved random walk image segmentation method is introduced to segment the reference image and the content image to obtain a more reasonable segmentation area,which can improve the layering of the result image.Secondly,use the color transfer strategy mentioned in part(1)to perform feature matching in the corresponding area to achieve color transfer.Finally,a structure-preserving filter is introduced to further optimize the result image to effectively improve the visual effect of the result image.A large number of experimental results show that the proposed method can achieve significantly better quality results than the color transfer method based on color clustering.(3)At present,there is no recognized evaluation index that can conduct a more comprehensive and objective evaluation of the color transmission effect.The existing objective evaluation methods are mostly carried out from two aspects of structure retention and color retention,so they can't fully evaluate the comprehensive effect of color transmission.In addition,some scholars use commonly used indicators such as peak signal-to-noise ratio,structural similarity,and feature similarity in image restoration to measure the degree of retention of the resulting image on the target image structure.However,practice has proved that the evaluation effect of these indicators is not ideal.In response to the above problems,the article combines color difference and improved structural similarity to design a comprehensive objective evaluation index suitable for color transmission.Experimental analysis shows that the newly designed comprehensive objective evaluation index for color transmission can effectively solve the one-sided problem of a single evaluation index,and can obtain evaluation results that are highly consistent with subjective evaluation.