Research On Image Colorization Based On Color Cues

Human beings observe things by vision and image and color are the two basic elements for them to perceive all things in the world. For human vision, the colorful image can express the objects lively, vividly and clearly. In the modern society, gray-scale images are still widely used in various fields of human life, but due to the recognition limitation of gray level by human eyes and the information lack of gray-level images themselves, colorful images include more extensive information and own more strong contrast compared to the gray-level images, which is more direct and easy to identify. It has important research significance in various areas to color the gray-level images and realize the colorful effect, especially in entertainment, medicine, remote sensing, animation and other fields. This paper applies the advanced computer technology into the color technology of gray level images and does a research on cue-based colorization technology and cue location method. The details are as follows.The paper first introduces the classical color techniques and digital image knowledge involved in algorithms, and then improves color blending algorithm that Yatziv proposed combining geodesic distance and unevenness. Beginning from the cue point, by the use of the brightness information of the gray-level image pixels, based on the calculation of geodesic distance and unevenness, it spreads the chromaticity throughout the whole image by color expansion, and then calculates the chromaticity of the final color pixel by combining the thought of chromaticity blending and weighting several chromaticity with maximum contribution. Then the paper will extend the color approach improved in the paper to the re-coloring issue of color images. As the luminance channel can only provide part of image information in color images, by the comprehensive utilization of the three channels of images, the paper use color distance to calculate geodesic distance of pixel, extend the new color provided by the users in the image space, and finally realize re-coloring of color images by color blending. The testing results show that the algorithm has high efficiency and good color effects.By researching the method of clue location, the paper proposes an image segmentation algorithm based on region growing for the color algorithm. The criterion of region growing is to expand from seed point to the neighborhood. The smaller geodesic distance based on brightness between neighborhood pixel and seed point is extended with priority. The research on clue method of clue location can effectively improve the efficiency of color treatment and ensure the adequate and rational coloring, thereby reducing the unreasonable color results.