Face image enhancement method based on multi-exposure face detection

Abstract

The invention discloses a face image enhancement method based on multi-exposure face detection. The method comprises the following steps: performing face detection on an image frame input into a facedetection model, obtaining a face position, segmenting the image into M*N regions, if the face position is detected, directly setting different weights for the M*N regions according to a designed algorithm, if the human face is not detected, judging a dynamic range of the current scene, and determining whether to switch exposure parameters or not according to the dynamic range, if the dynamic range is smaller than the preset threshold value, judging that no face appears in the current scene without switching of the the exposure parameters, setting the weights of the M*N areas as the default weights and if the dynamic range is greater than a preset threshold, switching different exposure quantities and exposure parameters according to a preset multi-exposure switching algorithm, and detecting the face under different exposure parameters.

Description

technical field [0001] The invention belongs to a face image enhancement method based on multi-exposure face detection, relates to the field of camera 3A and machine vision, and is specifically a method for improving the automatic exposure algorithm combined with machine vision face detection, thereby enhancing the face area . Background technique [0002] With the rise of machine vision, face detection algorithms have also achieved near-perfect results on existing data sets. However, a major feature of existing data sets is that faces are clear and exposure is reasonable; the existing human with the strongest comprehensive ability The detection effect of the face detection algorithm will be greatly reduced in harsh lighting environments, such as backlight, backlight, low light and other environments. Face detection in the field of machine learning is based on human facial features, with the corresponding classification algorithm learning model, so as to determine whether t...

AI Technical Summary

Problems solved by technology

In the scene where the lighting is uniform or the overall lighting is uneven but the default area of ​​interest is evenly illuminated and the face is just in the default area of ​​interest, the traditional exposure algorithm has a better exposure effect on the face, but under special lighting conditions, such as , the illumination is uneven, and the face is not in the default area of ​​interest, the phenomenon of abnormal exposure of the face area will appear, which not only has a negative impact on the subsequent machine vision, but also affects the direct human eye visual effects

[0004] The premise of the existing system method for intelligent exposure based on human face is to input a single frame image, detect the human face on the single frame image and modify the exposure weight table according to the position information of the human face to further obtain different exposure amounts and exposure parameters. The prerequisite for it to take effect is that the human face can be correctly detected in the input single-frame image. Therefore, once the human face is not detected during initialization, the subsequent algorithm cannot be carried out, and there is little information about the solution in this regard.

Images