Saliency detection in complex scenes

Linfeng Xu,Nii Longdon Sowah,Liaoyuan Zeng,Huiping Duan

doi:10.1186/1687-5281-2014-31

Abstract

Detecting multiple salient objects in complex scenes is a challenging task. In this paper, we present a novel method to detect salient objects in images. The proposed method is based on the general ‘center-surround’ visual attention mechanism and the spatial frequency response of the human visual system (HVS). The saliency computation is performed in a statistical way. This method is modeled following three biologically inspired principles and compute saliency by two ‘scatter matrices’ which are used to measure the variability within and between two classes, i.e., the center and surrounding regions, respectively. In order to detect multiple salient objects of different sizes in a scene, the saliency of a pixel is estimated via its saliency support region which is defined as the most salient region centered at the pixel. Compliance with human perceptual characteristics enables the proposed method to detect salient objects in complex scenes and predict human fixations. Experimental results on three eye tracking datasets verify the effectiveness of the method and show that the proposed method outperforms the state-of-the-art methods on the visual saliency detection task.

Highlights

Visual saliency is a state or quality which makes an item, e.g., an object or a person, prominent from its surroundings
We focus on a bottom-up model of saliency detection, which is widely believed to be closely connected to the ubiquity of attention mechanisms in the early stages of biological vision [24]
The saliency value is weighted by the radius of the saliency support region, which can be represented as

Summary

Introduction

Visual saliency is a state or quality which makes an item, e.g., an object or a person, prominent from its surroundings. These methods involve a variety of saliency models, such as biologically motivated (e.g., IT), computational (e.g., AC and MSS), frequency-based (e.g., SR), mixed (e.g., AIM and GB), local contrast (e.g., IT and AC), global contrast (e.g., RC), and state-of-the-art (e.g., SF, MR, and DSR) models.

Findings

Method

Conclusions