Human Hairstyle Generation Method Based on Multi-Feature Retrieval and Deformation

A human hairstyle generation method, based on multi-feature retrieval and deformation, comprising following steps:

• step 1:
• acquiring a hair style mask of a single frontal image of a human face;
• step 2:
• identifying feature points of the human face and match the feature points with a standard human face of a hair style database;
• step 3, aligning the single frontal image with the standard human face to acquire a corresponding hair region;
• step 4:
• calculating Minkowski distances between the hair region and hair masks of all frontal faces in the hair style database, and assigning corresponding weights after sorting the Minkowski distances from small to large;
• step 5:
• calculating flow direction fields of some of most similar hair styles for performing matching operation with a detecting hair style;
• step 6:
• training a deep learning network to detect hair styles of hair basic blocks at different scales, and matching the detecting hair style with a histogram of a plurality of candidate hair styles with a plurality of scales, to acquire different matching scores;
• and
• finally, taking out a most similar hair style picture.