An Attention-Driven Model for Grouping Similar Images with Image Retrieval Applications

Abstract

Recent work in the computational modeling of visual attention has demonstrated that a purely bottom-up approach to identifying salient regions within an image can be successfully applied to diverse and practical problems from target recognition to the placement of advertisement. This paper proposes an application of a combination of computational models of visual attention to the image retrieval problem. We demonstrate that certain shortcomings of existing content-based image retrieval solutions can be addressed by implementing a biologically motivated, unsupervised way of grouping together images whose salient regions of interest (ROIs) are perceptually similar regardless of the visual contents of other (less relevant) parts of the image. We propose a model in which only the salient regions of an image are encoded as ROIs whose features are then compared against previously seen ROIs and assigned cluster membership accordingly. Experimental results show that the proposed approach works well for several combinations of feature extraction techniques and clustering algorithms, suggesting a promising avenue for future improvements, such as the addition of a top-down component and the inclusion of a relevance feedback mechanism.

A Comparative Study of Image Retrieval Algorithms for Enhancing a Content-based Image Retrieval System

“…The main components of CBIR system are as follows [12]: 1) Graphical User Interface which enable the user to select the query which can an image example: content based image retrieval systems allow the user to specify an image as an example and search for the images that are most similar to it, presented in decreasing order of similarity score. 2) Query / search engine: it is a collection of algorithms responsible for searching the database for images that is similar to the user's query.…”

NLOOK: a computational attention model for robot vision

“…Thus, although it has some advantages related to NVT, SAFE has many restrictions, and thus is not adequate to be used in real time robotic vision systems. Other attentional models were also analyzed 11,13,29,34 , but their performance were similar to that of NVT in relation to the sensitivity to 2D similarity transforms. Thus, it was decided to propose and implement a new computational model of visual attention, called NLOOK, which is more adequate to be used in robotic vision task.…”

Using visual attention to extract regions of interest in the context of image retrieval

“…Images are then clustered together based on the features extracted from these regions. This process is detailed in our previous work [18]. The result is a group of images based not on their global characteristics, but rather on their salient regions.…”

“…The proposed method was inspired by the success of a recently developed computational model of the human visual attention [13] and is based on the knowledge of the salient regions within an image provided by such model. It is part of a new CBIR architecture -described in more detail in a separate paper [18] -in which these regions, once extracted, are then indexed (based on their features) and clustered with other similar regions that may have appeared in other images. This paper is structured as follows: Section 2 reviews relevant previous work in the fields of content-based image retrieval and computational modeling of human visual attention.…”