Colorization of grayscale image / Nuraiza Ismail
Colorization refers to an image processing task which recovers color of gray scale images when only small regions with color are given. This project introduces a methodology for adding color to grayscale images. Colorization represents a computer-assisted process of adding color to gray-scale still...
Saved in:
| Main Author: | |
|---|---|
| Format: | Student Project |
| Language: | en |
| Published: |
2008
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/43641/1/43641.PDF https://ir.uitm.edu.my/id/eprint/43641/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Colorization refers to an image processing task which recovers color of gray scale images when only small regions with color are given. This project introduces a methodology for adding color to grayscale images. Colorization represents a computer-assisted process of adding color to gray-scale still images or monochrome movies. Since the mapping from gray-scale pixel to color pixel is not unique, this process generally requires some user interactions to help confine the color selection.
The project presents a simple yet effective user-guided colorization method. In this method the user only required to annotate the grayscale image with the desired color scribbles in the interiors of the various regions. Two constraints are imposed, one as a constrained least-squares (CLS) which computed the intensity of neighbouring pixels. The other one is linearly constrained quadratic optimization (LCQO) that blends the computed color region in the chrominance images. These constraints are formulated as an optimization problem that automatically propagates the scribbled colors to produce a completely colorized image. The basic premise is that neighbouring pixels having similar intensities in the monochrome data should have similar colors in the chromatic channels. The colorization method requires neither precise manual image segmentation, nor accurate region tracking. |
|---|