Face image quality evaluation method and device and computer readable storage medium

Abstract

The invention discloses a face image quality evaluation method and device and a computer readable storage medium. The method comprises inputting a plurality of face images into a trained multi-task network model to obtain predicted values of a plurality of attributes corresponding to each face image, the plurality of attributes comprising a face angle, a face definition and face shielding information; and obtaining a face quality score of the corresponding face image according to the predicted values of the plurality of attributes. Through the design mode, the predicted values of the key features of the face image can be obtained in the same network in a targeted manner, and score quantification is performed on the face image. The efficiency of face image quality evaluation is improved, and computing resources and time cost are saved.

Description

technical field [0001] The present application relates to the technical field of face image quality evaluation, in particular to a face image quality evaluation method, device and computer-readable storage medium. Background technique [0002] With the development of digital society, image data plays an irreplaceable role in video communication, security monitoring and other fields. However, the image data in the video is inevitably disturbed in the process of collection, transmission and processing; secondly, in the video sequence, The image data collected for the same target will also have quality differences due to factors such as attitude angle and occlusion. In order to ensure the efficient use of video data in engineering applications, it is very meaningful to evaluate the quality of images. [0003] At present, commonly used methods include: method 1, quality scoring through quality scoring network learning, which combines image, cropping and multi-resolution quality...

AI Technical Summary

Problems solved by technology

[0003] At present, commonly used methods include: method 1, quality scoring through quality scoring network learning, which combines image, cropping and multi-resolution quality score output; method 2, combining the depth features corresponding to each face image with the same Comparison of standard human face features; however, in method 1, the quality score label is difficult to obtain, the quality evaluation features are vague, and the quality score network integrates pictures, cropping and multi-resolution quality score output, which is relatively redundant; in method 2, the operation process Not only need to define a standard face, but also need to compare the similarity between faces many times, which undoubtedly increases the time cost

Images