A new approach for image retrieval in search engines

Abstract EN

The ever increasing amount of images in WWW creates a need for new sophisticated methods to retrieve the visual content one is looking for.

In Content-Based Image Retrieval (CBIR), extraction of many features that express visual contents in image like color, texture and edges is necessary.

This paper is divided into two main parts, a first one concerns with the features extraction and the second concentrates on dimensionality reduction.

In the phase of features extraction, the 3D color histogram and the Gabor filter algorithm are used to extract the Color and texture features respectively.

The dimensionality reduction algorithm performs two simultaneous functions.

The first function (rough pruning) deals with the dominate values of feature which neglects the less information feature(s) via probability measure of the values inside each feature.

In the second function (fine pruning) the set of relevant feature is determined via the calculation of certain evaluation function.

This algorithm also presents a proposed search technique that reduces the search space linearly.

The main advantage of proposed algorithms is the balance between retrieval quality and retrieval time.