Research On Key Technologies Of Geometric Texture Synthesis And Image Resizing

Digital content industry is developing fast globally these years and playing a more and more important role in economy. Animated movie and3D game are very important parts of the industry, and they are both based on a key technology:rendering. A core process of rendering is texturing. In computer graphics, we can add textures to an object in two ways, with geometry texture and color texture. Geometry texture is added by modifying an object’s surface geometry, and color texture is added by altering an object’s surface color. For geometry texture, how to efficiently synthesis high quality geometry texture on surface is a difficult issue to solve. For color texture, as a special case of2D texture mapping, image resizing is a hot research topic in recent years.Based on these background, we study the geometric texture synthesis and image resizing problem. By synthesis regular geometric texture on an arbitrary surface, geometric texture synthesis can generate a new mesh. Image resizing technologies aim to protect important object from distortion when adjusting the resolution or ratio of an image. After systematically discussed the difficulties and current study on the two problems, we propose a new geometric texture synthesis method and two new image resizing methods. Our main works and contributions are:1. A geometric texture synthesis method based on Laplacian texture imageThe key idea is to use Laplacian texture images to represent geometric texture details, which in turn facilitates simple and effective geometry texture synthesis and enables flexible geometry texture editing. Given a sample model and a target model, we first select a patch from the sample model and extract the geometric texture details. Next, we construct a Laplacian texture image for the extracted texture and synthesize the Laplacian texture image to the target model using an image texture synthesis technique on surface. Finally, we reconstruct the textured target model with adjusted Laplacian coordinates. Compared to the existing methods, our method is easy-to-implement and produces results of high quality. Furthermore, it performs flexible control on the appearance of the textured target model through operations on the Laplacian texture image.2. An image resizing method based on spring analogyWe resize an image by representing an image as a triangular mesh in2D space and viewing each triangle edge as a spring. Through deforming the spring system, we can implicitly retarget the image to its new size. To build the spring system, we firstly run a saliency detection algorithm to generate each pixel’s saliency. Then we get the triangular mesh after running a Delaunay triangulation on mesh points distributed over the image according to saliency differences. Our deformation of the spring system involves solving linear equations only once, which is more efficient than existing warp-based image retargeting methods that use iterative solver. Finally, the original image is mapped onto the deformed spring system to get the retargeted image.3. An image resizing method based on Gaussian PyramidDuring image retargeting, how to keep significant image areas from obvious distortion is the main consideration. The seam carving method handles the challenge by recursively removing pixel seam with minimal cost. Though it can generate convincing results, the computation needed to find a seam is very time consuming. In this paper, we propose a new framework to speed up seam carving using Gaussian pyramid. First, we build Gaussian pyramid of an image. Then we distribute the seams to be removed over the pyramid layers. Removing one seam in the second layer means two seams in the first, and so on. Finally, we map seams in different layers back to the original image and remove them. In this way, we can reach good retargeting results while simultaneously reducing the number of seams to be removed and computational time for finding each seam.