Unsupervised Content based Image Retrieval at Different Precision Level by Combining Multiple Features

Abstract

Image retrieval is a procedure of finding appropriate images in the image database. There are two types of image retrieval systems in common practice. These are the text-based image retrieval (TBIR) system and content-based image retrieval (CBIR) system. The content based system is proven to be more effective in which the visual contents of the images are extracted and described by multi-dimensional feature vectors. In this work, several models are developed by combining different image features in a combination of two and three. To begin with, three different models based on the combination of two features, viz., color with shape, shape with texture, and color with texture are designed. A three features based model is considered with color, shape, and texture in the next step. The retrieval rate of the mentioned models is assessed in terms of precisions. The results are obtained using COREL standard database. This study shows that the images can be better retrieved using three features based model in contrast to models using two features.

Tasks

Reproductions