【作者】 王绘；

【导师】 邹北骥；

【作者基本信息】 中南大学 ， 计算机科学与技术， 2011， 硕士

【摘要】 人类通过视觉观察事物,而图像和颜色是人类通过视觉感知世界万物的两个基本要素,对于人类视觉来说,彩色图像能活泼、生动、清晰的表达事物对象,而当今世界灰度图像仍被广泛应用在人类生活的各个领域,但是由于人眼对灰度层次的识别限制和灰度图像自身的信息匮乏,相比灰度图像,彩色图像包含更丰富的信息,具有更强的对比度,给人的视觉效果更加直观和容易辨识。通过给灰度图像上色达到彩色图像的效果在各领域具有重要的研究意义,尤其是在娱乐、医学、遥感、动漫等领域,本文将先进的计算机技术应用到灰度图像的彩色化技术中,对基于线索的彩色化技术和线索位置提示方法进行了研究,具体工作如下：本文首先对经典的彩色化技术和涉及本文算法的数字图像知识进行了介绍,然后结合测地距离和不平度,对Yatziv的颜色混合彩色化算法进行改进,从线索点开始,利用灰度图像像素点的亮度信息对测地距离和不平度进行计算,通过颜色扩展将色度在整个图像中传播,最后结合色度混合的思想,将几种具有最大贡献的色度加权最终计算出像素点的色度。随后本文将改进的彩色化方法扩展,应用到彩色图像重着色问题上,基于亮度通道在彩色图像中只能提供一部分图像信息,因此本文综合利用图像的三通道,利用颜色距离来计算像素点的测地距离,将用户提供的新的颜色在图像空间中传播开,并同样的利用颜色混合的思想最终实现彩色图像重着色。试验结果表明,本文算法具有较高的效率,彩色化效果良好。对线索位置的提示方法进行了研究,本文提出了一种针对本文彩色化算法的基于区域生长的图像区域分割方法,区域生长的准则是由种子点开始向邻域扩展,邻域像素离种子点的亮度测地距离越小,则优先扩展。对线索位置的提示方法的研究能有效提高彩色化处理的效率,保证涂色的充分性和合理性,进而减少不合理的彩色化结果。更多还原

【Abstract】 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.更多还原