Re-colorization with random walks

• Main Author: Zhong, Yuan
• Other Authors: Cai Jianfei
• Format: Final Year Project
• Language: English
• Published: 2010
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/38566
• Institution: Nanyang Technological University

Image re-colorization means replacing the original image colors with other specifies colors. Image segmentation is the technique separating an image into different regions according to likelihood between adjacent pixels. My final year project conducted a research on combining random walks image segmentation and chrominance blending image re-colorization. The objective of this project is developing stand alone program realizing image re-colorization with segmentation performed first. The results are expected to be as good as the existing approaches. The program development was started after my solid understanding on the background knowledge. It took me around 6 months to finish the project. I used related papers and available source codes as the key references. I also adopted GMM in the random walks algorithm. The program developed gives quite accurate segmentation result, with natural re-coloring effects. Compared with best existing approach, it shows less sensitivity to users’ input. However, the speed of the program needs to be improved by various means. In the report, the introduction includes overview of the project, the project purposes, background information and limitations. Then it will discuss on design consideration as well as implementation details. After that certain test results will be presented to comment on program’s robustness. Last but not least, the conclusion wraps up the whole paper with meaningful recommendations on future works.