A Simple and Fast Region of Interest Extraction Approach Based on Computer Vision for Sport Scene Images

Region of interest extraction is very important in video analysis and understanding. A simple and fast region of interest extraction method for sport scene images is proposed in this paper. Firstly, a simple method is applied to detect the interest pixels in images by the defined interest pixels extraction function and a simple strategy is applied to improve the computation speed of mean for each pixel; Secondly, all the detected interest pixels neighbored each other are grouped to object interest regions; after that, the non-interest pixels enclosed in interest region are set as interest pixels by inner filling method; Thirdly, the binarized images composed by interest pixels and non-interest pixels can be easily segmented applying simple edge detection algorithms or contour tracing approaches; Lastly, several images captured from sport scene are used to test the proposed algorithm and the conclusion is given.