Eyebrow image segmentation method and system

Abstract

The invention discloses an eyebrow image segmentation method. The eyebrow image segmentation method comprises the following steps of acquiring a face image, processing the face image by using a cascade regression tree algorithm to obtain a plurality of ordered face feature points, respectively obtaining the candidate regions of left and right eyebrows according to the plurality of obtained orderedface feature points, extracting the images of left and right eyebrows from a data set by using the obtained candidate regions of left and right eyebrows so as to generate a new data set, and traininga total-convolution network by using the newly generated data set so as to obtain a well trained local eyebrow segmentation model. According to the invention, the candidate regions of eyebrows are determined firstly, and then the regions are subjected to exact segmentation. In this way, the operation efficiency and the robustness are well balanced. The technical problems in the prior art that, anexisting machine learning method is poor in robustness and an existing deep learning method is low in operation efficiency and slow in speed, can be solved.

Description

Technical field [0001] The invention belongs to the technical field of computer vision, and more specifically, relates to a method and system for segmenting eyebrow images. Background technique [0002] As an important aspect of information security, biometrics has attracted more and more attention. At present, the biometric recognition technologies that people research and use mainly include: face recognition, iris recognition, fingerprint recognition, voice recognition, etc. As an important feature of the human face, eyebrows have universality, uniqueness, stability and collectability as identifying features. [0003] The application of eyebrows for face recognition currently mainly includes traditional machine learning methods and deep learning methods. Traditional machine learning has the advantages of fast speed, but its robustness is poor; the advantages of deep learning methods are high robustness and high recognition accuracy, but its model is bulky, low in computational ...

AI Technical Summary

Problems solved by technology

[0004] In view of the above defects or improvement needs of the prior art, the present invention provides an eyebrow image segmentation method and system, the purpose of which is to balance the computational efficiency and robustness by determining the eyebrow candidate area and then accurately segmenting it , so as to solve the technical problems of poor robustness of existing machine learning methods and low computing efficiency and slow speed of existing deep learning methods

Images