RELIABLE POLYGONAL APPROXIMATIONS OF IMAGED REAL OBJECTS THROUGH DOMINANT POINT DETECTION

The problem of dominant point detection is posed, taking into account what usually happens in practice. The algorithms found in the literatu re often prove their performance with laboratory contours, but the sha pes in real images present noise, quantization, and high inter and int ra-shape variability. These effects are analyzed and solutions to them are proposed. We will also focus on the conditions for an efficient ( few points) and precise (low error) dominant point extraction that pre serves the original shape. A measurement of the committed error (optim ization error, E-0) that rakes into account both aspects is defined fo r studying this feature. (C) 1998 Pattern Recognition Society. Publish ed by Elsevier Science Ltd. All rights reserved.