Efficient Superpixel Construction Using Spatial and Color Information

• Main Authors: Wei-An Wang, 王偉安
• Other Authors: Chun-Rong Huang
• Format: Others
• Language: en_US
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/49736989027907604959

碩士 === 國立中興大學 === 資訊科學與工程學系 === 104 === We propose a novel superpixel extraction method to generate regular and compact superpixels. To reduce the computational burden of iterative optimization procedure of most recent approaches, the spatial and color quantizations are performed to represent pixels and superpixels. Then, pixels are assigned to the most spatially and visually similar superpixels using maximum a posterior estimation in the pixel and region levels. As a result, the extracted superpixels can precisely adhere boundaries of objects and are extremely efficient to generate. Experimental results show that the proposed method can achieve better or competitive performance compared to the state-of-the-art superpixel extraction approaches in terms of boundary recall, undersegmentation error and achievable segmentation accuracy and is significantly faster than these approaches.