Research On Perspective Texture Synthesis Technology

Texture synthesis technology is developed to solve the problems existed in texture mapping, such as seam, distortion and so on. The technologies of texture synthesis based on samples have gradually received extensive attention of the researchers, because of their ability to solve these problems well, and at the same time, to avoid cumbersome process of adjusting parameters. At present, texture synthesis has become the focus of study in the fields of computer vision, graphics and image processing, and computer graphics, etc. Perspective texture is a special category of textures, which usually possess perspective angle and the object distribution has certain pattern in scale. So synthesis of this type texture becomes difficult due to the need to keep these characteristics. On account of that perspective texture can give people a stronger sense of reality, as a result, people’s demand for perspective texture is bigger, the study on synthesis technology of perspective texture, in video games, film and television production, virtual reality, large-scale advertisement, computer-aided design and other industries, will have significant application prospect and value.On the basis of studying typical2D texture synthesis methods and related technologies, this paper puts forward an approach of combining image quilting and energy optimization technology to realize the synthesis of perspective texture. The traditional neighborhood-matching methods only pay attention to the local continuity of texture, but cannot meet its global feature. The global optimization approach, mainly represented by energy optimization, can realize the effective control to the global property of texture, thus has been widely studied. But in practical, energy optimization technique needs to compute each pixel, there exists the shortage that computation amount is large and synthesis time is very long. Using the thought from pixel-matching to patch-matching in texture synthesis for reference, the processing object in energy optimization is changed to be patch. At the same time, to avoid the problems like boundary overlap and seam because of patch processing, image quilting technique is applied into it, so that the traditional energy optimization approach is improved. The perspective texture synthesis algorithm combining image quilting and energy optimization techniques analyzes the input sample first, and calculates its corresponding scale map according to the angle given by user, then resamples the input scale map, according to the output size, to be output image corresponded scale map, and obtains the initial output texture by comparing the two scale maps, finally, applies the improved energy optimization technique to optimize and process the initial output. The experimental result shows that the proposed perspective texture synthesis algorithm combining image quilting and energy optimization techniques can effectively improve the image synthesis quality, and raise the computational efficiency.