Systems for segmenting an image based on perceptual information and methods for making and using same. According to one embodiment, input channels from an image are derived and analyzed by heuristic metrics to create categorical estimates. Examples of categorical estimates include a foreground channel estimate and a background channel estimate. Once created, the categorical estimates are merged to create a final channel estimate. The final channel estimate may represent the foreground or background of an image. Optionally, noise removal will also be conducted to improve the segmentation.