Exploiting Contextual Information for Image Re-ranking

Abstract

Abstract. This paper presents a novel re-ranking approach based on contextual information used to improve the effectiveness of ContentBased Image Retrieval (CBIR) tasks. In our approach, image processing techniques are applied to ranked lists defined by CBIR descriptors. Conducted experiments involving shape, color, and texture descriptors demonstrate the effectiveness of our method.

Efficient Image Re-Ranking Computation on GPUs

“…In the past few years, there has been considerable research on improving the distance measures in CBIR systems [1]- [3], [5]. Promising results have been obtained considering several approaches and techniques.…”

“…In general, traditional CBIR systems perform only pairwise image analysis, that is, they compute similarity (or distance) measures considering only pairs of images, ignoring the rich information encoded in the relations of several images. However, in recent years, several CBIR approaches [1]- [3], [5], [6] have been proposed to improve the effectiveness of retrieval tasks replacing pairwise similarities by more global affinity measures that also consider the relation among the database objects [2]. These approaches propose improving the effectiveness of image searches by exploiting the information about the relationships among collection images in an unsupervised way (with no training data).…”

“…These approaches propose improving the effectiveness of image searches by exploiting the information about the relationships among collection images in an unsupervised way (with no training data). Another important common point consists in the use of an iterative strategy adopted to process contextual information [1], [3], [6]- [9].…”

“…Context can be broadly defined as all information about the whole situation relevant to an application and its set of users. Promising results have been obtained [1]- [3] considering several approaches aiming at improving the effectiveness of CBIR systems.…”

“…We propose a parallel GPUbased solution which can speed up the Contextual Re-Ranking [1] algorithm computation.…”