How to generate a human hairstyle based on multi-feature search and deformation

摘要

Obtaining a hairstyle mask, identifying facial feature points and matching with a face in the hairstyle database, aligning an image with a standard face to obtain the corresponding hair area, and in the hair area and hairstyle database Calculate Minkovsky distances for all frontal facial hair masks, sort from smaller to larger, give corresponding weights, train deep learning networks and detect hairstyles in basic hair blocks on different scales How to generate a human hairstyle based on multi-feature search and transformation, including doing and retrieving the most similar hairstyles. In the present invention, it is possible to avoid constructing a model by hand by searching a large-scale three-dimensional hairstyle database for a three-dimensional hair model most similar to a photograph using one frontal face illuminator. Increased efficiency and guarantees higher fidelity. [Selection diagram] Fig. 2