A new type of video text automatic recognition method and its application in film and television works

Abstract

In the current post-production of film and television, editing and school teams need to identify each frame of image in the video and modify the sensitive words in it, which consumes a lot of manpower, material resources and financial resources. This article proposes an image recognition method based on attention mechanism to overcome this problem, which can quickly recognize specified texts in videos. Using the convolutional neural network and codec as the framework, the target text is used as the operation object of the attention mechanism, so as to realize the rapid recognition of pictures containing the target video text. To validate the proposed method, a TV series is taken as an example to test the characters involved in the video, and the results show the effectiveness of the proposed video text automatic recognition method.