An image retrieval technique using rotationally invariant Gabor features and a localization method

In order to search images in a database, a content based image retrieval methods are widely used for finding visually similar one to a given sample image. The approaches using Gabor functions and filters are regarded as one of the excellent methods for feature extraction and texture segmentations. However, the Gabor method does not perform well specifically for a rotated image since the filter has a direction oriented property. This paper proposes a method of extracting local texture features from blocks with central interest points extracted from an image and a rotationally invariant Gabor wavelet filter. A method of comparison among the pattern histograms that are constructed features are presented after the classification using vector quantization (VQ) between images.