闭合形式抠图算法比较及人像植入工具开发
毕业论文关键词：闭合形式抠图算法；比较；人像植入工具
Comparison of Close-form Image Matting Algorithm and Development of Portrait Implantation Tool
Abstract：The current mainstream image processing tool for automatic matting function of complex foreground edge, matting portrait prospect is not accurate, such as matting filament hair. In natural image matting algorithm the closed-form image matting algorithm can get the image edge details. Therefore, this paper chooses three kinds of closed-form matting algorithm: Sharing Matting, Improving Image Matting using Comprehensive Sampling Sets, KNN matting to study. The matting algorithm comparison has the following problems: (1) Comparison experiment did not compare different algorithm systematically and comprehensively from different conditions. The present study matting algorithm there are many difficult problems, such as the foreground color and background color are similar, the foreground edge is complex, so it is important to evaluate the matting algorithm in different scenes in closed-form matting algorithm comparison. (2) There might be several matting problem in one test image, thus local comparison can accurately illustrate the performance of the algorithm in solving a specific matting problem, there is no local compare the comparison experiments or no local quantitative comparison method of matting algorithm comparison. This paper presents a comprehensive evaluation scheme of closed form matting algorithm and selects three kinds of closed-form matting algorithm to evaluate. Firstly, this paper analysis of the algorithm's time complexity and space complexity in theory. And Standard test images representing different scenes and classified as the evaluation test set, this paper compare the matting algorithm in different scenarios of accuracy, stability and efficiency of the three aspects. The result is that shared matting algorithm is suitable to solve the problem of hair filaments prospects, and it has good stability and high efficiency among three algorithm. So this paper choose sharing matting algorithm to apply in the portrait implantation tool. Finally this paper develops a Portrait Implantation Tool, and it tests that the result is better than the current image processing tools.
Key words: Close-form Image Matting; Comparison; Portrait Implantation Tool
目 录
摘要 １
关键词 １
ABSTRACT １
KEY WORDS １
1 绪论 ２
1.1 研究背景 ２