Design And Realization Of No-reference Thangka Image Quality Evaluation System

With the advent of the technological information explosion,people need to get feedback from the outside world through various media every day.As a direct information carrier which is acceptable to human eyes,image has gradually become a more mainstream social tool.Thangka is a kind of image form with Tibetan artistic characteristics,its composition is rigorous and always contains many unique national culture,so it has become the historical relics and treasures of China.As time goes by,some Thangka images eventually be damaged irreversibly due to such factors as crease,mildew,repair and degradation caused by improper preservation or repair.Therefore,how to objectively evaluate the damage and repair degree of Thangka images has become one of the first problems to be overcome in the next cultural relic protection work.Considering that it is difficult to obtain the complete original information of some Thangka images due to objective factors such as history,this thesis mainly studies and discusses the quality evaluation method of damaged and repaired Thangka images without reference.(1)Aiming at the Thangka damaged and coloured image,this dissertation proposes a completely blind image quality assessment based on superpixel(DCTIQA-SP)combine color feature and human visual characteristics.First,the LAB color space feature method was used to extract the image with simple linear iterative clustering super-pixels.The pixels with similar features based on LAB color space were clustered into the same area,and the images were segmented based on color features.Then,the color entropy is used to quantify the color distribution features of the image,and the structure features of the damaged Thangka image are obtained by mean filtering of the color entropy.Finally,the quality evaluation score of the image is calculated.In order to solve the problem that the score is not objective due to the small amount of data in the damaged Thangka database,the algorithm carried out the operation of image inversion from different angles,and the mean score is taken as the final evaluation index.Experiments show that Pearson and Spearman correlation coefficient in this paper are improved,which is more consistent with the subjective evaluation results.(2)Aiming at the symmetrical Thangka inpainting image,this thesis proposes a completely blind inpainting image quality assessment based on harris corner point(STIIQA-HCP).This method takes into account that most Thangka images are symmetrical,especially Buddha images,and uses it to predict reference images.Then,an image quality evaluation model is constructed by extracting Harris interest points and quantifying the difference between the predicted and damaged area.The experimental results of Thangka images show that the algorithm can obtain better results than the reference image quality evaluation method without the original image data,and can effectively overcome the difficulty of image repair quality evaluation with complex edge.(3)Finally,aiming at the problem that there is no complete Thangka image quality evaluation system at present,this thesis uses MATLAB to design and implement a quality evaluation system for Thangka damaged and repaired images,which can achieve image segmentation,image rotation,feature extraction and acquisition of different image quality evaluation scores.A lot of tests and experiments shown that the system can achieve the goal set at the beginning of the design.