The recognition of patterns is accomplished through boundary tracing and subsequent storage of encoded range testable data commensurate with the occurrence, interrelationship and orientation of geometric features on the pattern boundaries. The encoded data is compared with generalized prototypes which define the geometric shape of all probable permutations of all possible patterns to be recognized, a match between the encoded data and a prototype constituting identification of the pattern.
In a picture encoding and decoding method and apparatus, an object contour can be chain coded precisely even in the case where noise exists in the vicinity of object contour, and the object contour branched off can be coded with small quantity of codes. In the case where the branched part occurs in a feature point chain, by multiplexing branch code D1 to chain codes S23 and S22 and expressing as a tree structured feature point chain including the branch part of the object contour, an increase in number of chains can be controlled and the efficient chain coding can be conducted. Moreover, since the terminal branch chains having no branch point in the chain become the objects of thresholding, only noise elements can be effectively eliminated without cutting off the feature point chain showing the object contours.
In a picture encoding and decoding method and apparatus, an object contour can be chain coded precisely even in the case where noise exists in the vicinity of object contour, and the object contour branched off can be coded with small quantity of codes. In the case where the branched part occurs in a feature point chain, by multiplexing branch code D1 to chain codes S23 and S22 and expressing as a tree structured feature point chain including the branch part of the object contour, an increase in number of chains can be controlled and the efficient chain coding can be conducted. Moreover, since the terminal branch chains having no branch point in the chain become the objects of thresholding, only noise elements can be effectively eliminated without cutting off the feature point chain showing the object contours.
A known hand drawn stroke is serially sampled at different scales by n angle filters to provide n filtered known strokes at n different scales. An unknown hand drawn stroke is sampled in a like manner to provide n filtered unknown strokes at n different scales. Filtered unknown strokes are compared with filtered known strokes. If there is a comparison within predetermined limits at any scale, the unknown hand drawn stroke is recognized as being the same as the known hand drawn stroke.
One or more contours of a picture element (PEL) array pattern are detected and, then, a strength of concavity/convexity on each contour (acuteness) and a direction of concavity/convexity are determined. Using an acuteness value of a contour point of interest, its curvature is classified into a group of strong curvatures or another group of weak curvatures such that the definitions of a direction of the concavity/convexity (contour direction) is appropriately switched from one to another for calculation. In a preferred aspect of this invention, using an acuteness value of a contour point of interest, its contour curvature is classified into either one of five curvature groups, comprising "strong concavity," "weak concavity," "straight line/inflection point," "weak convexity," and "strong convexity." If its contour curvature is classified into either "strong concavity" or "strong convexity," then calculation of a normal line direction will be made for deriving a contour direction. On the other hand, if its contour curvature is classified into "straight line/inflection point," then calculation of a tangent line direction will be made. In so doing, it is possible to improve image recognition rates compared to those of the prior schemes utilizing conventional feature values.
A feature recognition method and apparatus is described where in one embodiment the digitized two-dimensional pattern or image usually of a character to be recognized is subjected first to conversion into a vector. The vector subsequently undergoes propagation and stopping operations subject to various tests so as to generate a list of unique features about the pattern being identified. Thusly, identified features are compared mathematically to a previously created feature library until a match is found. In an alternative embodiment, to increase the speed of three-dimensional, feature-extraction processing of character patterns the vector-extraction positions on a pattern are converted into flags and accommodated in a corresponding motion-flag memory. Flag values in the motion-flag memory and a stopping-flag memory are read sequentially after propagation processing, and it is determined whether there are flag values representing the existence of vectors in the flag values in the above addresses of other flag memories and addresses adjoining thereto. The flag values in corresponding addresses in a self-moving flag memory are reset, and at the same time, the flags are set in corresponding addresses in the stopping-flag memory relative to vectors with the same direction, if there are vector flag values in the corresponding addresses. The flag values in each stopping-flag memory and each motion-flag memory are converted into vector values via determination processing after propagation processing, and are output as three-dimensional features.
