Research And Implementation Of Image Color Editing With GPU Acceleration

With the rapid development of the Internet, especially the mobile Internet, people’s demand for information is increasing. With the vigorous development of digital imaging technology, visual information such as image, video and other visual carriers has become an important source of information for people to obtain information. People are increasingly strong demand for image processing. As one of the most popular subject of image processing, image color editing has a wide range of applications.Manifold preserving edit propagation is a kind of linear dimension reduction algorithm of image editing. The main idea is using the local linear embedding algorithm to ensure that the linear combination between one pixel and the adjacent pixels while color editing. Due to the increasing of image pixel level, the processing time of existing image color editing algorithms is too long. Based on this algorithm, we use the GPU parallel architecture to improve the algorithm, so as to achieve higher operation efficiency.Our work mainly has two aspects:(1) Improve the KNN method of the algorithm for finding nearest neighbors of pixels. Using CUDA’s multi-thread, high-parallel characteristics to move the calculations of the distances between data and the sorting process to GPU, which greatly reduced the running time (2) The calculation process of manifold structure after dimension reduction is a sparse matrix equation solving process, we use GPU to accelerate sparse matrix vector multiplications. Then we achieve the Jacobi iteration method for solving sparse matrix equations on GPU, and improve the algorithm speed.In order to verify that our method can accelerate the image color editing, we set up some experiments to verify the feasibility of using GPU to speed up the image color editing. Experiments demonstrate that our method can improve the speed of image color editing.