Sketch-based perception model for evaluating color segmentations
David San José Gavilán, Hiroki Takahashi, Suguru Saito, Masayuki Nakajima · 2005
Global purpose segmentation in uncontrolled environments, usually based on human perception, has gained attention in recent years [Comaniciu and Meer 2002]. In this work, we propose an evaluation method to measure how much close are the segmentation results to human's perception. This study contributes to the image processing field by adding pictorial knowledge to segmentation algorithms, and to image retrieval applications by aiding in the selection of a segmentation method.