Optimizing image normalization algorithm for shape distortions

摘要

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 drawbacks of the normalization algorithm, shape compacting proposed by J. G. Leu, which normalizes rotation and skewing distortions incompletely, an optimized shape normalization algorithm is proposed in this paper. The basic idea is first to get the compact shape which is invariant to translation and scaling distortions by the shape compacting. Then, on determining the principal axis of the object shape, we get the angle included between the x-axis and the principal axis, according to which the shape is rotated. Finally, the reversed object shape can be normalized by the signs of the original image's central moments. Therefore, we can normalize a shape and its distorted versions into a single one, with the following feature descriptor invariant to the above four distortions. The results of our experiments demonstrate that the optimal shape normalization algorithm outperforms the existing shape compacting.