Recognition of Rotated Images Using Singular Value Decomposition Applied to Complex-log Mapping

概要

論文の詳細を見る
In the field of digital image recognition, the extraction of rotation-invariant features is left as an important problem. In this paper, the recognition of rotated images is made by applying singular value decomposition to complex-log mapped images, which are generated as pre-processing from original images. Fourier transform has conventionally been used for the shift-invariant method to the mapped image, but singular value decomposition is at this stage employed. That is to say, singular value decomposition is executed for the mapped image, and eigenvectors and singular values of the mapped image are calculated. For the translation of the mapped image adjustment of the eigenvectors can obtain a list of the same positive singular values. In utilizing the method using the singular value decomposition, just a small number is sufficient for the features extracted from the mapped image. Thus high efficiency accomplished. In case, for example, that the number of the pixels of the mapped image is N × N pieces, the number of the features is no less than N pieces with singular value decomposition. Finally effectiveness in relation to recognition of rotated images is shown by conduction computer experiments. At this stage, the recognition ratio is computed by applying the error-back-propagation learning to a 3-layer hierarchical neural network with one hidden layer.
東海大学の論文