Methods for Texture Feature Extraction in Image Retrieval

This paper introduces a texture feature extraction method based on Gabor filters and Gabor wavelet transforms. The method analyzes images to extract texture features and applies Gaussian normalization to Gabor wavelets to enhance both the speed and accuracy of image retrieval. Specifically, the approach begins by applying Gabor filters to process images through convolution operations with carefully selected frequency and orientation parameters. The Gabor wavelet transform then extracts multi-scale texture features by decomposing images into different frequency bands. Subsequently, Gaussian normalization is applied to the wavelet coefficients to standardize feature distributions and improve retrieval performance. Through this methodology, we can more accurately capture texture information from images, thereby enhancing both the quality and efficiency of image retrieval systems. The implementation typically involves configuring Gabor filter banks with multiple orientations and scales, followed by statistical analysis of normalized wavelet coefficients for feature vector generation.