Research On Digital Restoration Of Cultural Collection Images

China is an ancient civilization country with a long history of 5000 years.In the development process from ancient times to the present,the ancestors left countless precious wealth.However,due to the changes of the times,many historical relics have been damaged to varying degrees.How to repair them quickly and effectively has always been an urgent problem for human civilization.With the continuous improvement of science and technology,the form of repairing the digital images of damaged cultural relics through modern digital restoration technology and then saving them in the digital museum has gradually become the mainstream of the development of cultural relics protection.In this paper,the main research object is the digital image of cultural collection.Through the in-depth analysis of the corresponding repair mechanism of different types of defect images,a comprehensive repair system based on image segmentation,color migration and partial differential equation is proposed.The research content mainly includes the following aspects:In view of the color and texture features of the image are not taken into account in the traditional gray scale histogram segmentation algorithm,this paper proposed a new color image segmentation algorithm based on Lab sub channel histogram to solve the problems such as over segmentation or under segmentation for no obvious gray value difference images or grayscale overlapped images.The algorithm introduces three segmentation bases with sequence irrelevance: luminance L channel,red-green a channel and blue-yellow b channel,and then use the Newton interpolation method to fit operation.Users can choose the channel freely for different brightness and chromaticity attribute images.The paper also uses the matching principle of adjacent area gray value to solve the exact matching problem of pixels in the adjacent target area.The algorithm realizes the extraction of different objects in the image by sub office,sub morphology and subarea.In the light of images with significant difference in regional luminance characteristics or regional color difference greater than regional brightness difference,the segmentation results of the algorithm in this paper are better than the traditional gray scale histogram segmentation algorithm through a series of experimental verification,it greatly improves the applicability of histogram segmentation algorithm.A region self-adaptive image segmentation model based on multi-channel auto-theory is proposed.Firstly,using the uncorrelation theory of Lab sequence to eliminate the disadvantages of single target area affected by multi-channel interaction;Aiming at the defect that the segmentation effect of classical segmentation model is unsatisfactory when dealing with a color image whose membership degree is not obvious,the multi-channel auto-theory is introduced to determine the optimal screening threshold "m" by weighting the "standard deviation" and "maximum deviation" in the autologous samples of the region to be segmented.The channel with the greatest difference between the attributes of the area to be separated and the background area is selected as the main body of the segmentation system.Then,according to Welsh adaptive theory,the improved adaptive segmentation model which takes into account the pixel color appearance,8-neighborhood mean and8-neighborhood standard deviation is adopted to realize the pixel-by-pixel matching between the source image and the reference screening sample.Then,the multi-group matching results of each pixel of the source image are counted,and the threshold "m" is used to filter and distribute the statistical results to achieve the purpose of segmentation.Finally,in view of the problems of over-segmentation and under-segmentation in the operation process,mathematical morphology theory is used to refine them.The experimental results show that the proposed algorithm can effectively separate target regions from different types of background environments,and verify the superiority of the improved model in segmentation accuracy compared with the classical model.A gradient-preserving color transfer algorithm based on multi-channel adaptive matching is proposed to improve the classical Welsh algorithm and Reinhard algorithm.L,a and b channels are introduced as matching criteria,the processing of a single channel will not affect the other two channels because of the sequence irrelevance between channels,eliminating the drawbacks of single target region in RGB color mode that is vulnerable to multi-channel influence.For images with similar gray values,the channel with the largest difference in the mean chromaticity between the processing area and the background area is adaptively matched with the corresponding channel of the reference image,which solves the problem that Welsh algorithm can easily misjudge the grading of color source images with small gray differences or overlapping gray ranges.Then,gradient factor is introduced and weighted with the standard deviation ratio of reference image to source image.The result is taken as the scaling ratio coefficient of the improved transfer algorithm in this paper.Gradient-preserving color transfer algorithm is used to color each group of matched pixels to avoid the problems of loss of color detail information and unnatural transition between tones in Reinhard algorithm.Experiments show that the color effect of the improved algorithm is better than that of the classical Welsh algorithm and Reinhard algorithm,which greatly improves the flexibility and applicability of the color transfer algorithm.An improved BSCB restoration model based on adaptive segmentation system is proposed.Firstly,the region to be repaired and the intact region are separated accurately by the region self-adaptive image segmentation model based on sub-channel auto-theory,and then the region to be repaired is designated by Reinhard color transfer algorithm for automatic computer recognition;Finally,the improved BSCB model with ASG operator is used to repair the marked area.The max/min function is used to selectively extract some neighboring points,and the appropriate weighting algorithm is used to make the model better maintain the edge and corner points when repairing the image with large defect area.This algorithm solves the problem of image blur and isolux line crossing caused by Laplacian smoothing operator considering the same-sex diffusion of all neighboring points in the process of traditional BSCB algorithm.Experimental results show that the improved algorithm has better repair effect for different defect types of source images.In this paper,the restoration mechanisms of different types of defects of cultural collection images are deeply studied,such as fading,discoloration,damage,smudgy,wrinkle and so on.By combining the restoration scheme of Lab sub channel histogram segmentation technology and Reinhard color transfer technology,the problem of "Significant difference between target and background color" image restoration is solved;By combining the region self-adaptive image segmentation technology based on the sub-channel auto-theory and the improved BSCB technology based on ASG operator,the problem of "Non significant difference between the target and the background color appearance" image restoration is solved;By using the gradient preserving color migration technology based on the sub-channel pixel by pixel matching,the problem of "The whole situation is full of defects" image restoration is solved.A complete set of digital image restoration system for cultural collection has been formed.The experimental results verify the feasibility and universality of the system,which provides strong support for the inheritance of China’s long history and culture.