Texture Synthesis Algorithm And Its Application In Texture Transmission

For any object of the nature world,texture can describe it’s complex surface detail. Texturesynthesis is based on the existing texture information, synthesizing a large area of a similartexture,and this technology has been more and more attention in recent years, and it has becomean important research in the field of image processing now. Texture synthesis technology isdesigned to meet the people’s requirements for synthetic texture images, and it has been widelyused in photorealistic and non-photorealistic rendering, for example it has broad applicationprospects in the image restoration, image art style transformation, rapid transmission ofcompressed network data and computer animation, etc. With the rapid development of computertechnology and human society, people are increasingly demanding high digital image quality,which requires texture synthesis technology development should continue to progress andrelevant research staffs keep the exploration and innovation to realize human pursuit of higherdigital images..This article firstly introduces the research background, research significance and status quoof texture synthesis, followed by presentation of the related theories and basic concepts,andanalysis of several classic texture synthesis algorithms and classic algorithm for certain summary,a general description of the advantages and disadvantages of the algorithms. On this basis, toimprove texture synthesis picture quality and accelerate the speed of texture synthesis,this papermakes corresponding improvements to texture synthesis algorithm based on image quilting in therespect of cutting path, matching block selection strategy and block searching policy. Theimproved algorithm results compared to the previous algorithm synthesis results, both in termsof speed or synthetic synthesis quality have made great improvement. Finally, this paper appliesthe improved texture synthesis algorithm to the texture transfer process, and achieve the imageartistic style conversion experiments quickly and efficiently, and explore the transmissionconstraints measurements on transmission process of the two images,and in different colorspaces, uses different metrics as the constraints to make corresponding experiments. The mainfindings are as follows:First: For the reason of image quilting texture synthesis picture at the seams prone togenerate excessive rough phenomenon, the original algorithm is improved by tracking anddetecting the cutting points, calculating gradient of each cutting point of the cutting path,marking the pixel of greater gradient value than a predetermined threshold value as rough cuttingpoint,smoothing the marked over rough sharp points and exporting them again in the cutting pathby the way-weighted refinement.Second: In order to improve the efficiency of matching block selection, namely to improvethe quality of the texture synthesis image, the choice of matching block matching strategy hasbeen improved, the above improvements of cutting path in the right worth way to join to thechoice matching strategy,which makes that the strategy considers both the overlap regioncumulative error and the minimum cutting path when selecting the matching blocks. Theimproved matching strategy greatly improve the similarity of synthetic images and sampleimages.. Third: To reduce the synthesis time and increase the speed of texture synthesis, textureblocks for global search methods are proposed to improve by the new way of index tablesearching based on the texture of local relevance. The improved algorithm is according to thatsynthesized texture block mapps in the sample image as well as according to the neighborhoodcorrelation of the mapping texture block to make indexing table, using the established table tosearch matching block to achieve the matching process. This search way than the global searchmethod greatly reduces the search matching time, improving texture synthesis speed.Fourth: The above modified image quilting texture synthesis algorithm is applied to thetexture transmission, realization of the graphic arts style conversion through experimen.In adifferent color space, exploring texture transmission constraint metrics, through experimentproving texture transmission constraint measure is diverse, not a single, this diversity of theconstraint measure provides a richer choice for users to get more artistic style of images.