Hair strand extraction by template matching using Gabor filter and histogram

Hair modeling is an essential part for many applications in computer graphics. Hair modeling starts with viewing hair as individual strands. In this work with an image-based technique we extract two dimensional hair strands to pass them (as input) to another system which generates their 3-D hair model. The key idea is that working on small sub-images is easier than using the whole image at once to extract strands. We break down the input hair image to blocks and replace each with the matched template from created manual database. The method uses Gabor filter to find the orientation map of the pixels in the input image and templates in database. A similarity measurement is applied to find the best match. In next step we make connection between blocks to create strands.