Tuesday, 23 November 2021

Assessment of Automatic Hair Colorization and Relighting Using Chromaticity Distribution Matching | Chapter 12 | New Visions in Science and Technology Vol. 9

 This study outlines a novel method for colourizing and relighting human hair. It's difficult to colour human hair while maintaining a certain model's hair picture without modifying the hairdo or texture. The variation in hair texture and illumination between a user and model image is the basic challenge that makes this operation difficult. Human hair is made up of a variety of hair swatches. Each swatch has its own chromaticity distribution, which is non-Gaussian in most cases. These swatches are treated as colour clusters in the hair image by the suggested approach. Matching the user's and model's hair swatches or colour clusters fixes the problem in this scenario. The colour transfer between the relevant model and user swatches is then applied after this matching. Furthermore, the model's hair should be compressed to an acceptable size in order to portray a variety of hair hues at the same time. The model's hair colours were inspired by photographs of hair colour packs found in most ornamental beauty stores. These photographs, on the other hand, were shot in ordinary lighting circumstances, therefore relighting is required to get a photorealistic user appearance. The suggested technique achieves great photorealistic perception and a fair compression ratio, according to experimental findings with 530 different colour models and over 20,000 users. A high pick signal to noise ratio (39 dB) implies that the difference between original and recreated model hair colour is barely discernible.


Author(S) Details

Uri Lipowezky
Research and Development Department/Facetrom, Tel Aviv, Israel.

View Book:- https://stm.bookpi.org/NVST-V9/article/view/4791

No comments:

Post a Comment