Analysis on Extraction of Common Foreground Object by Co-Segmentation

Co-segmentation is a subclass of segmentation used to segment out the common objects by assigning multiple labels to segment the targeted common things. Methods existing so far do not exhibit competitive results for any specific task. The key issue of the proposal selection-based co-segmentation problems lies in mining consistent information shared by the common targets. Due to uncertainty of the shared features, it usually requires manually selecting features or feature learning performed beforehand. The main goal is the comparison of an existing and proposed co-segmentation method. Proposed method (second method) focuses on reduction of running time. For each image, a multi-search strategy extracts target individually. Experiments are orchestrated on public dataset MSRC and iCoseg. Performance evaluation of both the methods is compared.