Research And Application Of Coloring Methods For Bilateral Grids

With the rapid development of the Internet,photographic equipment has become more and more popular,and a large number of pictures have emerged every day.Pictures have become an extremely important carrier for transmitting information,so people's demand for image quality has gradually increased.When the directly photographed image does not meet the needs of people,people often use image editing to edit the image.Therefore,image editing algorithms have gradually become a research hotspot in related fields,and image coloring methods have received more and more attention in recent years.Coloring is one of the important research directions in the field of image editing.Originally,coloring was a computer-aided processing technique used only to color black and white images and video.With the development of science and technology,black and white video and images gradually disappear,and coloring can be extended to re-color color video and color images.This paper first introduces the existing coloring methods,and briefly introduces the three methods of coloring based on user interaction,coloring based on reference image and coloring based on machine learning,and analyzes some of the classic algorithms in detail.Because the existing methods generally have the problem of time inefficiency,this paper proposes a coloring algorithm based on bilateral grid which can improve the time efficiency.The algorithm is mainly divided into two steps:image preprocessing and color editing propagation.The main contents are as follows:(1)This paper first re-samples the input image with user interaction using a bilateral grid,so that all pixels in the input image fall into the corresponding bilateral grid.(2)Using the user interaction pixel in the bilateral grid as interpolation,the radial basis interpolation function is used to construct the interpolation model and optimize it.(3)Coloring the target bilateral mesh vertices to be colored according to the optimized interpolation model,and inversely mapping the coloring results to the pixels included in the target bilateral mesh to obtain a final coloring result image.The experimental results show that the proposed algorithm significantly improves the time efficiency on the basis of ensuring high quality effects.The main research object of this paper is image coloring.The target image of image coloring contains a large number of natural images with repeated scene elements and natural images with complex backgrounds.Therefore,how to effectively process these two types of images is the current research difficulty.The method proposed in this paper is applied to these two types of images and compared with other three different coloring methods.The experimental results show that the proposed method has small calculation and high time efficiency on the basis of obtaining excellent coloring results.