An Improvement And Application To The Image Quilting Texture Transfer Algorithm

For the most of the objects, we can use the surface texture to describe the intricate details.Texture synthesis technology has received extensive attention from the beginning to the present and has application in many fields, it is a hot spot in the field of computer graphics, computer vision and image processing. In recent years, the texture synthesis from samples is widely concerned. Along with the texture synthesis from samples, the defects of texture mapping and the procedural texture synthesis are effectively changed, and bring a broad space for the texture synthesis.This paper based on the original image quilting texture transfer algorithm by Efros et. al., reached the producer on the texture synthesis and texture transfer, and modify the error matching formula in the process of texture transfer, and taking the luminance remapping method for sample image and target image, at last we take the image similarity measure for the effect figure of improved texture transfer algorithm and discusses the problem of the diversity of constraints in the process of texture transfer. The research work of this paper mainly includes:1. Summarized and analyzed the development history of the texture synthesis, and take on a detailed introduction for the classical algorithm of texture synthesis.2. The luminance remapping method for the sample image and target image. We use the luminance mapping method which proposed by Hertzmann, Luminance remapping method is introduced as a preprocessing stage to diminish the matching texture block searching ghost caused by too large luminance difference between the texture and the target images.3. Modify the error matching formula of image quilting texture transfer algorithm, and add a term relating to the edge gradient information of the target image. Firstly, check the edge information for the target image, and get together the edge information as a collection edge. Secondly, when the texture synthesis is beginning, check the collection edge, if it has the edge gradient information and extracting the information for error calculation, if not, then the information is 0.4. Take image similarity measure. This paper takes two methods for measure the similarity of images, calculate the difference between the corresponding pixels and level histogram matching gray. We compared the results of the image quilting algorithm and the new texture transfer algorithm for the difference between the corresponding pixels, and the smaller the difference is, the higher the similarity of the two images.5. Explore the difference effect for the different transmission constraints of the texture transfer results. We can controlled by the different texture transfer constrain from the experiment, and respectively control the different color space, the different color components and the varying from 0 to 1 for the R components of RGB color space. From the different experiments we can get different artistic style images, it is provides great support for the different artistic style images from texture transfer.Through a lot of experiments, the new texture transfer algorithm for this paper proposed is superior in transmission performance, and it is also proved the diversity of the transmission constraints for texture transfer, and can get the different artistic style texture transfer images from different transmission constraints.