【作者】 于俊清； 杨明； 何云峰；

【Author】 Yu Junqing+, Yang Ming, He Yunfeng (School of Computer Science & Technology, Huazhong University of Science and Technology, Wuhan, 430074, China)

【机构】 华中科技大学计算机科学与技术学院；

【摘要】 精彩镜头提取是足球视频处理与检索技术的一个重要组成部分。针对目前现有的足球视频精彩镜头提取方法存在运算量过大、可控性差和精彩镜头类型固定的问题,提出了一种基于精彩度的足球视频精彩镜头提取方法。精彩度分为声音精彩度和视觉精彩度两类,首先通过声音平均短时能量和声音平均过零率特征计算声音精彩度,获取可选精彩镜头,然后在可选精彩镜头的基础上,依据关键帧分类计算视觉精彩度,确定最终的精彩镜头。实验结果证明该方法可以有效地提取足球视频的精彩镜头,并具有很好的伸缩性。更多还原

【Abstract】 In recent years, the processing and retrieval of soccer video is increasingly rapid developing and has become a hotspot. Although soccer video last for a long time, there are only a few highlights which the viewers just want to watch directly inside it. How to exactly extract highlights is an important step in the research of soccer video. Based on the analysis of massive computation and the fixed highlights type of the existed extraction methods, a highlights extraction technique, named Exciting Degree-based Highlights Extraction Method, is proposed. Exciting Degree includes audio exciting degree and visual exciting degree. Audio exciting degree is computed by the ratio for the number of the frames whose features meet the needs in the whole soccer video. The highlights are followed by a few scenes, for example the short scene and the player scene. The visual exciting degree is summation of the frame type value for these shots. The proposed approach computes audio exciting degree using the features of audio short-term energy and zero crossing rates to extract the candidate highlights, and uses frame classification to compute the visual exciting degree for highlights verification. Experimental results indicate that the proposed method extracts the highlights efficiently. However, only a few types of exciting degree are designed for audio and visual information. How to accurately calculate the audio exciting degree and the visual exciting degree is our future research emphasis.更多还原