Abstract—This paper presents an efficient framework for key events detection and summarization in cricket videos. The proposed research work presents a non-learning technique based on textual features to detect three key events in cricket videos that are, boundary (4), six (6) and wicket. Image averaging is used to detect the score captions from the input video that is then analyzed to detect changes in score and wickets counters. To extract the contents of score captions, input video frame is discretized by using the mean and standard deviation. Morphological operators are applied to get rid of the noise and outliers. The extracted score caption region is passed to the optical character recognition algorithm to analyze any significant change in the score and wicket counters. The frame is marked as a key frame in case any significant change (boundary, six, or wicket event) is detected. A collection of video frames are selected against each key-frame to generate the summarized video. The proposed method is tested on a diverse dataset of cricket videos belonging to different tournaments. Experimental results illustrate the efficiency of the proposed method for key events detection and summarization of cricket videos.

Index Terms—erosion, key events, morphology, optical character recognition, score captions, video summarization

Cite: Muhammad Haseeb Nasir, Ali Javed, Aun Irtaza, Hafiz Malik, and Muhammad Tariq Mahmood, "Event Detection and Summarization of Cricket Videos," Journal of Image and Graphics, Vol. 6, No. 1, pp. 27-32, June 2018. doi: 10.18178/joig.6.1.27-32