Face depth image quality evaluation method and system, electronic equipment and storage medium

Abstract

The invention provides a face depth image quality evaluation method and system, an electronic device and a storage medium, and the method comprises the steps: obtaining three-dimensional face recognition image data based on an input target image, and the three-dimensional face recognition image data comprises aligned face depth images and corresponding identity category information; inputting the aligned face depth image and the corresponding identity category information into a face depth image quality evaluation model to generate a corresponding face depth image quality score; wherein the face depth image quality evaluation model is obtained by training based on face depth image sample data and a predetermined face depth image real quality score label. According to the invention, end-to-end face depth image quality evaluation is realized, and the face depth image quality is comprehensively evaluated; the three-dimensional face recognition image is screened through the face depth image quality result, it is ensured that the face depth image quality meets the three-dimensional face recognition requirement, and therefore the improvement of the three-dimensional face recognition performance is promoted.

Description

technical field [0001] The invention relates to computer image processing, video monitoring, face image quality evaluation and convolutional neural network neighborhood, in particular to a face depth image quality evaluation method and system, electronic equipment and storage medium. Background technique [0002] In recent years, face recognition technology has been successfully applied in various fields, and the 3D face recognition system that uses face depth images for recognition has also attracted much attention. The face recognition system has been able to achieve a high level of user cooperation and controllable environment. However, most of the practical application scenarios of face recognition are uncontrollable, resulting in a significant reduction in the quality of the acquired face images, thus reducing the performance of the face recognition system. The recognition effect of a face recognition system is affected by image acquisition conditions such as illuminati...

AI Technical Summary

Problems solved by technology

This type of method only considers a limited number of attributes that may affect the quality of the face depth image, and cannot comprehensively evaluate the quality of the face depth image. The quality evaluation module of the face depth image is independent of the face recognition module, and the obtained better quality Face depth images are not necessarily easy to identify for face recognition systems, and cannot effectively reflect the impact of face depth image quality on face recognition performance

[0004] Some methods based on supervised learning can comprehensively evaluate the quality of face color images. First, the real quality score label of the training data must be obtained. Generally, the real quality score label is obtained by artificially scoring the face image or selecting a benchmark image. However, artificial The evaluation results may not be fully applicable to the subsequent face recognition algorithm. In addition, the workload of manual selection is relatively large, and the human consumption is relatively large.

For face depth images, there are no more mature algorithms to generate reliable ground truth quality score labels

Images