On optimizing image normalization algorithm for shape distortion

In general, there are four basic forms of planar shape distortion caused by changes in viewer's location: rotation, scaling, translation and skewing. For a good shape descriptor should be invariant to these distortions, a shape can be normalized before feature extraction. Due to the problems of the normalization algorithm, called shape compacting, proposed by J. G. Leu, which cannot deal with rotation and skewing distortion thoroughly, this paper proposed an optimized image normalization method. The basic idea is to get the compact image first which is invariant to translation and scaling distortions by the shape compacting algorithm of Leu. In a second step, we find the main axis of the object shape and the angle between the x-axis, then rotate the image according to the angle. At last, we deal with the reversed object shape if possible by the sign of the original image's central moments. This algorithm can normalize a shape and it's distorted versions (rotated, scaled, translated, skewed) into a single result, so that they all become highly similar to each other. Therefore, the following feature descriptor will effectively be invariant to the above four distortions. Experimental results show that the improved algorithm is correct and effective, outperforms the existing shape compacting method