Study On Exemplar-Based Thangka Image Inpainting Technologies

With the rapid development of digital image processing technology, the connection between the individual life and the digital image is becoming more and more closely. Thangka image is a very precious image of Tibetan features, and then some thangka image was damaged due to improper preservation, so repair the broken thangka image has very important significance. The first chapter of this paper introduces the basic concepts, research background and application fields of digital image inpainting technology. Based on this, the second chapter introduce the technology of image inpainting based on local feature of image, image inpainting based on block matching and image inpainting technology based on spatial fusion.Summed up the current based on the sample piece of thangka image at the same time still exist deficiencies in inpainting technology, and at the same time for in-depth analysis of these disadvantages.Based on study the problems of the sample piece on image inpainting technique in inpainting thangka image, such as the mistake matching phenomenon and the inaccurate matching standards, the third chapter of this paper puts forward the image inpainting algorithm based on thangka image structure information. The algorithm is mainly combined with the characteristics of thangka image, first of all conversion to the image to Lab space to repair, the main priorities for improvement in the process of repair, make the priority update more accurate, improve the matching standards, make the match more accurate, reduce the false match, make the structure of the image more coherent after inpainting.Based on the statistics of observating a lot of thangka image, we found that there has duplicate information in many tangka images, and the thangka image texture is more complex. For this kind of thangka image, we propose an image inpainting algorithm that combined image direction characteristic structure matching standards, the algorithm is through the two inpainting to achieve the inpainted image. The purpose of the first inpainting for the broken thangka image is computing eight direction factor, preparing for the second inpainting process. In this paper, the main improvements are manually select matching area, this greatly save the inpainting time, and update the priority to reduce the mistake matching diffusion phenomenon, using the direction factor and the color of the Euclidean distance to combine a new matching criterion. Analysis the influence of hand-choose matching area size on the experimental results, the impact of false match for the results of the experiment. Experiments show that this algorithm after repair of thangka image, edge coherent structure is smooth, false match than comparison algorithm is significantly reduced.In order to increase the practicality of the algorithm, this paper designed a visualization of thangka image inpainting system combined with MATLAB GUI platform, make the inpainting of thangka image more convenient.