Fig. 8
From: Understanding and simulating border ownership centered segmentation
Random Dot Stereograms (RDS) Dataset, generated using the modified method from Marino (2007)28. Top row: Left and right RDS images, where the left image is a pure random dot image, while the right image is created from the left image by shifting random numbers (1 to 6) of randomly located rectangle-shaped areas. The shifted amount of each area is randomly selected from values of 2, 4, 6, 8, 10 and 12 pixels; Middle row: The left image is a ‘Near’ disparity map, where darker red rectangles indicate closer objects (larger shifts), and lighter red rectangles indicate farther objects (smaller shifts). The right image represents a ‘Far’ disparity map, which is the logarithmic form of the ‘Near’ disparity map. Bottom row: The left image shows a 2-channel border ownership map for the ‘disparity-defined’ objects; the right is an anaglyph of the two RDS images.
