Method for annotating subtitles in news video

Abstract

The invention discloses a method for annotating subtitles in a news video. The method comprises the following steps of: gathering the news video; decoding the news video to obtain single-frame video images; extracting candidate subtitle areas from the video images to locate news subtitle areas; matching the news subtitle areas with template images in a news template base; if the news subtitle areas are matched with the template images, annotating the news subtitle areas according to the annotation formats of the template images; and if the news subtitle areas are not matched with the template images, annotating the news subtitle areas. By adopting the technical scheme disclosed by the invention, the accuracy of annotating the subtitles in the video can be increased, the artificial check time is reduced, and a news video retrieval effect is enhanced.

Description

technical field [0001] The invention relates to the technical field of video content analysis, in particular to a method for tagging news video subtitles. Background technique [0002] With the development of the information age and the increase of video data, it is increasingly showing a massive scale. Compared with other types of content, video content has a more vivid display form and carries richer information. In order to achieve efficient access to video information and facilitate people to efficiently browse and retrieve it, the development of video processing and retrieval tools has become a top priority. Video is a combination of image, sound, and text. The subtitles in the video are an important source of high-level semantic content of the video, such as the title of the news, content prompt information, broadcast time, narration, thank you and work after the program Personnel lists and the like contain rich high-level semantic information. Extracting subtitle i...

AI Technical Summary

Problems solved by technology

Such marking is not conducive to manual proofreading and retrieval of video information, and of course it is not conducive to readers' browsing of these news videos

[0006] Manual labeling can ensure the accuracy of information, but the sales volume of manual labeling is low, the cost is high, and the information that can be extracted is limited

At the same time, with the rapid growth of data volume, this method makes labeling become unacceptable

The computer program automatically locates and recognizes and then assists in manual proofreading. In the video subtitle extraction method, the subtitle area can be quickly located based on the connected domain, but it is sensitive to the background complexity; the texture-based algorithm is simple to implement, but due to the influence of similar texture areas in the video. The accuracy of detection is low; the edge-based method has low computational complexity, but it is very dependent on the background of video subtitles; there are also some methods based on statistics to build classifiers, which are too dependent on training samples

Existing video subtitle processing methods have their own problems, and it is difficult to find a method that can deal with the current news video situation

Moreover, the existing system does not make full use of the characteristics of news videos. It only considers a general video subtitle annotation. The general type is not good and has not considered the different types of subtitles in the video.

As a result, the effect of automatic positioning and recognition of video subtitles is not ideal, resulting in an increase in the workload of manual proofreading, which increases the cost of labeling

Images