Abstract
This paper presents a 2D shape retrieval system using moment invariants based on a representation of image boundary of the shape objects. The original binary image is transformed into a matrix which consists of numbers that represents the distance to the shape contour. Chessboard distance transform is used for the representation of the shape contour. Moment invariants are computed for the original binary image and the proposed transformed contour image. In this paper, we use radial basis probabilistic neural networks to recognize the object shapes based on moment invariants features extracted from the shape contour. The effectiveness of the proposed shape retrieval system using moment invariants based on image contour is reported.