Semantic Image Retrieval by Combining Color, Texture and Shape Features

The volume of digital images generated and uploaded on the internet every day by the scientific, medical, educational, industrial and other communities are very large. The problem of retrieving the desired images from huge collections is a major problem. The user queries are becoming very specific and traditional text-based methods cannot efficiently handle them. The subjectivity of human perception and the rich contents of the images further aggravate the problem. To overcome this problem, a new query-by-example technique using multiple color, texture and shape features is proposed and evaluated in this paper. The experimental results suggest that our proposed technique is efficient and retrieves semantically more similar images.

I. Introduction

In the last few decades there has been enormous growth in the field of internet as well as digital multimedia. Many methods have been developed in the literature to store and manage these kinds of data. Content based image retrieval (CBIR) system analyses image contents via the low-level features for indexing and retrieval, such as color, shape and texture. In order to achieve higher semantic performance, these systems seek to fuse low-level features with high-level features that contain perceptual information of human beings [1]. However, such fusion increases the feature extraction time and the memory requirement as well as the retrieval complexity. CBIR, also known as query by image content, does not consider human-input metadata such as the keywords and texts, whereas query by text considers it. CBIR system is composed two main steps. First step extract various features of the images and store in a feature database. Second step match the feature of query image with the features stored in the feature and find the images which are visually similar to the input image [2].