Abstract: A hierarchy of border attribute data items is produced from a body of data such as the pixel values of an image. Each border attribute data item indicates an attribute of a border of an analysis region of the image. The starting border attribute data items, at the lowest level of the hierarchy, are produced by operation on the pixel values. Each border attribute data item at a higher level in the hierarchy is produced by operating on a respective set of data items from the next lower level. The border attribute data items can indicate, for example, the number of black pixels in each border or the number of edge pixels in each border. Or the border attribute data items could include boolean values indicating whether any of the pixels in a border is a black pixel with a neighboring black pixel outside the border.

Abstract: And in-place hierarchical technique shifts a spatial processing focus within a two-dimensional array in a SIMD parallel framework. The hierarchical process uniquely associates a rectangular region with each image location at each level of the hierarchy. Each of these regions at a given level is the union of two child regions at the next lower level, one of which is associated with the same image location. In an upward pass, an occupancy bit is stored at each location at each level of the hierarchy, indicating whether the respective region at that level includes any pixels that meet a specified criterion for selecting the new processing focus. In a downward pass, the focus is shifted at each step to a pixel associated with one of the child regions of the region associated with the focus, the child region's occupancy bit being on. This involves a random or arbitrary choice between the two children when both of their occupancy bits are on.