Superpixel based Foreground Extraction in Surveillance Videos

• Main Authors: Chia-Chun Lee, 李佳駿
• Other Authors: Chun-Rong Huang
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/9t9hy2

碩士 === 國立中興大學 === 資訊科學與工程學系 === 103 === In the paper, we aim to extract foreground objects from surveillance videos based on superpixels. To obtain superpixels, simple linear iterative clustering (SLIC) [1] and saliency map detection (SMD) [2] are used to extract candidate regions of background superpixels and foreground superpixels, respectively. Then, we consider the spatial- temporal relationship and the visual similarity of candidate regions to extract the foreground objects. In the experiments, we compare our method with the Gaussian mixture models (GMM) [3] and saliency map detection (SMD) [2], and show the effectiveness of our method.