Database of human segmented images and its application in boundary detection

This study presents a database that can be used for boundary detection and image segmentation evaluation. Images in our database are based on image pairs instead of single images, which is a remarkable difference from existing databases. This characteristic will greatly facilitate the algorithms combining low-level cues such as brightness, colour and texture with depth information generated from the image pairs to conduct boundary detection and image segmentation tasks. Besides, the authors have put the presented database to use by comparing their proposed boundary detection method with the Berkeley detector and Edison detector. Experimental results show that their method has a better performance.