基于特征点的电影字幕提取

Movie Subtitle Extraction Based on Feature Points

电影字幕作为电影图像的解释文本,可以帮助观众更好地理解电影剧情.针对在复杂背景下提取字幕困难的问题,提出基于特征点、颜色等多特征融合的电影中英文字幕同时提取算法.该算法首先通过帧间差分算法进行运动目标检测,提取电影图像中的字幕帧;然后通过运用ORB特征点检测技术、启发式筛选、形态学处理等方法定位字幕,对准确定位的文本区域通过颜色聚类提取字幕文本,实现文本分割;最后通过OCR文字识别技术实现对字幕帧文本的识别.实验结果表明:该方法将处理帧数减少到与字幕条数相同,将字幕提取的准确率提高到95%,提高了字幕检测效率,验证了多特征融合技术的有效性.

As the explanatory text of movie images,movie subtitle can help audiences to better understand the plot of a movie.To address the difficulty of extracting subtitles in complex backgrounds,a simultaneous extraction algorithm for Chinese and English subtitles in movies is proposed based on the fusion of multiple features such as feature points and colors.In this algorithm,the moving target is detected by inter-frame difference algorithm,and the subtitle frame in the movie image is extracted.Then,by using ORB feature point detection technology,heuristic filtering,morphological processing,and other methods to locate subtitles,color clustering is used to extract subtitle text from the accurately located text area,achieving text segmentation.Finally,OCR text recognition technology is used to recognize the subtitle frame text.The experimental results show that this method reduces the number of frames processed to the same as the number of caption bars,the accuracy of subtitle extraction was improved to 95%,improves the efficiency of subtitle detection,and verifies the effectiveness of multi-feature fusion technology.