Face sketch synthesis method and system based on gabor features

Abstract

The invention provides a GABOR characteristic based face sketch synthetic method and system. The synthetic method comprises the following steps: dividing a face photo to be synthesized, a face photo sample and a face sketch sample into mutually overlapped pixel blocks; extracting the Gabor characteristics of each pixel block, and acquiring the Stein divergence matrix on the basis of the covariance matrix of the Gabor characteristics of the pixel block; acquiring the optimal weight on the basis of the Stein divergence matrix and the rebuilding coefficient of each pixel block of the face photo sample; adopting the optimal weight for weighting synthesis on the pixel block at the corresponding position of the face sketch sample to acquire a synthetic face sketch pixel block; integrating the synthetic face sketch pixel block to acquire the synthetic face sketch corresponding to the face photo to be synthesized. The synthetic method and system can synthesize the face sketch in higher quality, and can be applied to criminal investigation.

Description

technical field [0001] The invention belongs to the technical field of human face sketch synthesis, and in particular relates to a Gabor feature-based human face sketch synthesis method and system. Background technique [0002] In recent years, retrieving photos of criminal suspects from face photo databases has been widely used in law enforcement as a means of criminal investigation. However, sometimes the photos of the suspects cannot be obtained, and the artist can only draw a sketch of the suspect's face based on the memories of the witnesses, and then use the sketch of the face for retrieval. However, due to the obvious heterogeneity of face sketches and face photos, and the unknown psychological mechanism involved in the generation of face sketches, the accuracy of recognizing face sketches using traditional face recognition methods is very low. One of the methods to solve this problem is to synthesize pseudo-face photos or pseudo-face sketches based on face sketches ...

AI Technical Summary

Problems solved by technology

[0012] The existing sketch synthesis methods all use the Euclidean distance to measure two images or image blocks. Since the illumination has a great influence on the pixels, the high-frequency information will be lost due to the illumination change, and the synthesis effect will drop sharply. Therefore, the synthesis effect of the above method needs to be To further strengthen, it is also necessary to find a feature space that is more robust to illumination

Images