Defining Quantization Strategies and a Perceptual Similarity Measure for Texture-Based Annotation and Retrieval

In this paper, we introduce an approach for texture-based annotation and retrieval. Given the outputs of 12 Gabor filters, we derive a texture feature space where the sensitivity of the features to illumination changes is attenuated by a suitable normalization. We then annotate images by defining and selecting codes representing the quantized levels of the texture features appearing in each image. The annotations are stored in a hash table for retrieval efficiency. Ranking schemes are proposed to order the images retrieved at query time. In particular, we use results from psychological studies on the human perception of similarity to formulate a similarity measure. The choice of quantization of the texture feature space can influence the accuracy of the retrieval. We have compared several quantization schemes in retrieval experiments involving texture images. We have found that a uniform quantization and a quantization heuristically taking the variance of the texture features into account lead to the best retrieval performance.