Local auxiliary-color maximum vector pattern: a new feature descriptor for image indexing and retrieval

Abstract

A new feature descriptor, local auxiliary color maximum vector pattern (LACMVP) has been proposed in this paper for image indexing and retrieval. The presented method synergize the color and texture information by taking a cardinal (red, green, blue) and an auxiliary channel (value) from two different color spaces (RGB and HSV). A vector pattern comprising of magnitude, sign and position patterns are calculated for the maximum local difference between the center pixel and its neighbor from the auxiliary channel. In essence LACMVP converts the image into local vectors along the maximum edge of the inter-chromatic texture pattern. The performance evaluation of proposed method has been done by performing natural image retrieval on Corel-10K and texture retrieval on MIT VisTex dataset. The results when compared with existing state-of-the-art techniques using standard performance evaluation measure like precision, recall, F1-score and G-score, showed a substantial improvement.