Video segmentation method and device

Abstract

The present application discloses a video segmentation method and device. The method includes: a global feature analysis step, a local feature analysis step, and a shot segmentation step, wherein the global feature analysis step calculates each frame of the video based on the color space The first similarity between a frame, according to the first similarity to determine the candidate starting frame of the new shot; the local feature analysis step calculates the descriptor of the key point in the candidate starting frame and its previous frame to the visual word The distance value corresponds the descriptor to the visual word, respectively constructs a visual word histogram, and calculates the second similarity; the shot segmentation step judges the second similarity, and merges and confirms the candidate starting frame. This method combines global feature analysis and local feature analysis. While using global features to measure shot changes, it further refers to the local features of the picture to solve the problem of inaccurate shot segmentation.

Description

technical field [0001] This application relates to the field of image processing, in particular to the fields of automatic video image processing and the like. Background technique [0002] A shot is a continuous shooting process of the camera, which represents a continuous action in time or space. A video is composed of frames, a group of consecutive frames can constitute a shot, several consecutive and related shots constitute a video scene, and a group of semantically related video scenes constitutes a basic narrative plot segment. It can be seen that the lens is in a very basic and important position in this hierarchical relationship. With the development of multimedia, digital TV, and network video applications, the demand for video data structuring arises as the times require, and finding an effective algorithm to segment videos is a basic requirement for follow-up processing such as video analysis and retrieval. . [0003] Common lens segmentation methods are gener...

AI Technical Summary

Problems solved by technology

However, no matter which method is used, whether the RGB color space or the HSV color space is used, it is limited to the global characteristics of the picture, and cannot reflect the similarities and differences of the detailed features of the picture, and the effect is not good in the application.

Images