Pattern recognition by local radial moments

Identification of images irrespective of their location, size or orientation is one of the important tasks in pattern analysis. Use of global moment features has been one of the most popular techniques for this purpose. The authors present a simple and effective method for image representation and identification which utilizes local radial moments of segments of image as features as opposed to global features and a simple classifier such as the nearest-neighbor classifier. The technique does not require translation, scaling or rotation of the image itself. Furthermore, it is suitable for parallel implementation and hence is useful for real-time applications. The classification capability of the technique is demonstrated by experiments on scaled, rotated and noisy images of upper and lower case characters and digits of the English alphabet.