Colorization of monochrome images with texture

This paper examines the various colorization techniques that have been invented and improvised in the recent decades. The main issues with the techniques are the computational costs, poor and unrealistic quality of the colorized results, and the intensive user intervention required. In this project, Levin et al. [8]’s colorization program was implemented and the results analyzed to ascertain if the implementation of Horiuchi and Kotera [6]’s algorithm can resolve the textural flaws of the result. Horiuchi and Kotera [6]’s method involves the blending of the color of the seeds sowed in the monochrome image. This was based on the assumption that pixels with similar luminance and pixels which are near to each other should have similar colors. The normalized Euclidean and luminance distance between each pixels are thus used to determine the color of the monochrome pixels. A GUI was also built to integrate both programs together for effective user interaction and to create the same ground for comparison. The drawbacks of Horiuchi and Kotera [6]’s proposed algorithm were studied and suggestions were given to enhance and increase the robustness of the program. Future explorations are possible with the progress of colorization developments.Bachelor of Engineering (Computer Engineering