Systems and methods for unsupervised local boundary or region refinement of figure masks using over and under segmentation of regions

A system for labeling a plurality of pixels in an image comprising processor and a memory, the memory storing

• (A) means for generating an initial figure mask estimation of the image using a figure ground segmentation system, thereby initially assigning each pixel in the image with a first attribute value or a second attribute value;
• (B) means for generating a JigCut region segmentation of the image, wherein the JigCut region segmentation comprises a plurality of JigCut regions;
• (D) means for utilizing at least one computer to process the figure mask estimation with the JigCut region segmentation by
• classifying each pixel in a JigCut region in the plurality of JigCut regions with the first attribute value when a first predetermined number or a first predetermined percentage of the pixels within the respective JigCut region have been initially assigned the first attribute value by the generating (A), and by
• classifying each pixel in a JigCut region in the plurality of JigCut regions with the second attribute value when a second predetermined number or a second predetermined percentage of the pixels within the respective JigCut region have not been initially assigned the first attribute value by the generating (A),
• thereby generating a second figure mask estimation of the image in which the first attribute value is assigned to each pixel in a first set of pixels in the plurality of pixels and the second attribute value is assigned to each pixel in a second set of pixels in the plurality of pixels;
• and
• (D) means for outputting the second figure mask estimation to a user interface device, a monitor, a tangible computer-readable storage medium, a computer-readable memory, or a local or remote computer system;
• or displaying the improved figure mask estimation.