Determining Object Orientations from Run Length Encodings

Abstract

Run length codes are widely used for image compression, and efficient algorithms have been devised for identifying objects and calculating geometric features directly from these codes. However, if the image objects are rotated it can be difficult to determine their orientation and position so that they can be grasped by manipulators. This paper describes a method for structural determination of object orientation directly from the run length codes of successive image scan lines. An algorithm is described that makes use of the equations of object boundary segments to form hypotheses about object orientations that are refined as scanning progresses. 2-dimensional polygonal objects are discussed, and it is assumed that objects do not touch or overlap, although the algorithm could be extended to include those situations.