Research On Multi Spectral Image Digital Inpainting Technology For Painting Arts Reproduction

As a treasure of human civilization,especially the painting and calligraphy of ancient dynasties,traditional Chinese painting and murals and other works of art,Painting Art has immeasurable artistic value.However,there are different degrees of damage because of the human,accident,natural environment and other factors.Multispectral image digital inpainting for artworks reproduction is aimed at obtaining objective and accurate artwork color information by using multispectral techniques,which also used for color image digital inpainting work,digital collection of art,high fidelity image reproduction and so on.It can meet people’s demands on artwork color accuracy and structural integrity.The technology of traditional images obtains chroma information based on RGB three-channel,it can’t characterize the color information of the Painting Art objectively.This method is limited.So it will inevitably lead to inaccurate color inpainting and is prone to the appearance of pseudo-color if used for digital image inpainting.Therefore,this thesis studies the multispectral imaging technology and its application in color digital image inpainting.Multi-channel images are acquired by a multispectral imaging system.We can get multispectral images with more rich and true color through image registration,filter noise,reconstruction and rendition.Digital inpainting using this color multispectral image not only can get rich colors of inpainting image,but also can get inpainting quality better than RGB image which acquired by traditional method.The method can provide a new way of technology for Painting Art inpainting,and also provide a new idea of image acquisition and digital inpainting for Painting Art high fidelity image reproduction.(1)This thesis discusses the theoretical framework of digital inpainting technology for Painting Art based on multi-spectrum and expounds the key problems.The thesis analyzes the limitations of traditional image acquisition technology,which inevitably leads to pseudo-color in image inpainting.It cannot meet the demands on color accuracy and inpainting accuracy of Painting Art.It can obtain the color information of Painting Art accurately and can be more accurate for digital repair based on multispectral image.And then it can meet the demands of high fidelity reproduction on color accuracy and structural integrity.Here we discuss the key technologies which are involved in this thesis.(2)The thesis analyzes the advantages of multispectral images in color reproduction,high fidelity image reproduction and discusses the principle of spectral imaging,and studies the multi-channel image spectral reconstruction technique based on training samples.We propose the spectral reconstruction algorithm based on principal component and polynomial regression method for nonlinear imaging system,and obtain a certain improvement in spectral reconstruction accuracy and chroma accuracy.(3)In this thesis,the selection method of training samples in the process of spectral reflectance reconstruction is discussed,and the limitations of traditional training samples on sample selection are analyzed——the spectral characteristics and chroma characteristics of the color cannot be comprehensively reflected,and the similarity of the training samples is not good.We propose a training sample selection method based on minimum correlation of spectral spatial and similarity clustering in chroma spatial,which takes into account the spectral characteristics and chroma characteristics of the sample comprehensive.Experiments show that spectral reconstruction accuracy has been improved.(4)This thesis discusses digital image inpainting principles and main theoretical methods,and studies the main digital image inpainting technology and solves the phenomenon of pseudo color caused easily by traditional image inpainting technology.Therefore the inpainting precision is improved greatly.This thesis analyzes the inpainting problem of damaged Painting Art image for high fidelity images reproduction.The digital inpainting method of color image is analyzed from three aspects.And we propose an optimization method to solve the pseudo-color phenomenon caused by inpainting of color image.The thesis proposes an improved algorithm based on digital image decomposition and color image inpainting flow.We put forward to image structure inpainting method based on KL transform to inpaint each channel of RGB individually considering the pseudo-color caused by inpainting inconformity of each channel because of the RGB channel relevance.The inpainting result is ideal and appearance of pseudo-color reduces in the image repair quality.We also propose a method to calculate the image illumination line in three-channel irrelevant HIS color space which is suitable for color image analysis and detection to image texture information.It is difficult to accurately define the illumination line because of the RGB channel relevance.Experimental results show that RMS、PSNR、SSIS are improved,multispectral image inpainting effect is better,and it is suitable for inpainting work of Painting Art high fidelity image reproduction.It lays the foundation on multispectral image inpainting.A complete multispectral color digital image inpainting framework for art reproduction is proposed.And we validate the effectiveness of the inpainting framework through practice.It improves the accuracy of color image acquisition and the accuracy of image inpainting for damaged Painting Art.It lays the foundation on high fidelity image reproduction.