Improvement of Zernike moment descriptors on affine transformed shapes

In general, Zernike moments are often used efficiently as shape descriptors of image objects, such as logos or trademarks that cannot be defined by a single contour. However, because these moments are defined in a unit disk space and extracted by a polar raster sampling shape, information of skewed and stretched shapes is lost. As a result, they can be inefficient shape descriptors when there is skew and stretch distortion. In this paper, a method is proposed that addresses this issue. More specifically, Zernike moments are obtained from a transformed unit disk space that allows for the extraction of shape descriptors which are invariant to rotation, translation, and scale as well as skew and stretch, thus preserving more shape information for the feature extraction process. The experimental results demonstrate that the proposed algorithm is more accurate in relation to skew and stretch distortions when compared to other available schemes reported in the literature.