Video streaming face detection method and device

Abstract

The invention relates to the field of video processing, and discloses a video stream human face detection method and a device thereof. In the present invention, the following steps are included: obtaining the current frame image; determining the zoom scale range required when the current frame image is zoomed according to the first face window probability value of the zoomed image of the previous frame image, and performing the zoom scale within the zoom scale range Perform zoom processing on the current frame image to generate an image pyramid of the current frame image; generate a detection window for each zoomed image in the image pyramid; determine whether the generated detection window contains a face, and count the detection windows containing a face in each zoomed image and the number of detection windows that do not contain faces, the probability value of the first face window of each zoomed image of the current frame is obtained for use when the next frame image is zoomed. The invention can reduce the scale range of the image zooming according to the statistical value, reduce the calculation amount of the image detection, improve the detection speed of the human face, and save the calculation resources of the system.

Description

technical field [0001] The invention relates to the field of video processing, in particular to the face detection technology of video streams. Background technique [0002] Face detection is a basic research content in computer vision and pattern recognition. It is the premise of face recognition and has a wide range of applications in the fields of human-computer interaction, video retrieval and intelligent monitoring. The purpose of face detection is to find the position and size of the face in the image or video stream. At present, the most commonly used method for face detection is the AdaBoost (an iterative algorithm) algorithm proposed by Paul Viola (Paul Viola) and Michael Jones (Michael Jones). advantage. [0003] The AdaBoost algorithm uses Haar-like features to form a cascade classifier, which is obtained through training. When performing face detection, in order to detect faces of different sizes, the original image is scaled into a set of images to be detecte...

AI Technical Summary

Problems solved by technology

[0004] The traditional AdaBoost algorithm has many redundant calculations in the calculation, and it needs to consume a lot of computing resources and time to detect all windows of each zoom scale, especially for video streams.

Images