Video scene change detection method based on self-adaptation threshold value

Abstract

The invention discloses a video scene change detection method based on a self-adaptation threshold value. The video scene change detection method based on the self-adaptation threshold value comprises the steps that the similarity coefficient between color histograms of every two adjacent image frames in a video file is calculated; all the similarity coefficients are sequentially connected to form a similarity curve; a sliding window is set, the self-adaptation threshold value of the similarity curve in the sliding window is determined, the maximum value of the similarity curve in the sliding window is found out and the position, corresponding to the maximum value, in the video file is recorded; if the maximum value of the similarity curve in the sliding window is larger than the self-adaptation threshold value determined according to the sliding window and the number of the image frames spaced between the position, corresponding to the maximum value, in the video file and a former video scene change position is larger than a preset value B, the position, corresponding to the maximum value, in the video file is judged as a video scene change position. Through the video scene change detection method based on the self-adaptation threshold value, disturbance caused by abrupt change of a shooting angle or shot object can be well avoided, detection is rapid and accurate, and high application value can be achieved.

Description

technical field [0001] The invention relates to a video image analysis technology, in particular to a video scene change detection method based on an adaptive threshold. Background technique [0002] Content-based video processing includes analysis of video structure, automatic indexing of video data, and video reorganization. The analysis of the video structure is to divide the video into the basic unit - the shot by detecting the boundary of the shot; the automatic indexing of the video data is to select the representative frame from the shot, and use its characteristics as the representative of the shot feature; the reorganization of the video includes Scene extraction and stitching of multiple video segments enabled by features representing frames. [0003] Shot refers to the content obtained by the camera in one continuous shooting, and it is the basic unit of video. Shot cut refers to the transition from one shot to another shot, through the detection of shot cuts, m...

AI Technical Summary

Problems solved by technology

[0007] 1) When processing video, the color histogram features are extracted for all video frames, and the entire image is scanned to count the number of pixels with each color level. This will increase the complexity of the entire algorithm and affect Video processing speed

[0008] (2) When extracting the color histogram, each pixel is scanned for the entire video frame, without considering the spatial position information of the pixel in the video frame, which will cause some shot boundaries to be missed in the shot segmentation result

[0009] (3) When determining the boundary of the gradient lens, it is necessary to calculate the frame difference between the frames, which will also increase the computational complexity of the algorithm

[0010] (4) Changes in lighting conditions and flashlights will cause changes in the brightness of the video frame, causing changes in various video features, so it is easy to be mistakenly detected as the boundary of the lens

[0011] Therefore, the existing video scene change detection methods have the defects of large amount of calculation, complex processing, and low detection accuracy.

Images