Saliency

Saliency Detection: A Spectral Residual Approach

Xiaodi Hou and Liqing Zhang

Department of Computer Science, Shanghai Jiao Tong University

No.800, Dongchuan Road, Shanghai

http://bcmi.sjtu.edu.cn/˜houxiaodi, zhang-lq@cs.sjtu.edu.cn

Abstract

The ability of human visual system to detect visual

saliency is extraordinarily fast and reliable. However, computational modeling of this basic intelligent behavior still

remains a challenge. This paper presents a simple method

for the visual saliency detection.

Our model is independent of features, categories, or

other forms of prior knowledge of the objects. By analyzing the log-spectrum of an input image, we extract the spectral residual of an image in spectral domain, and propose a

fast method to construct the corresponding saliency map in

spatial domain.

We test this model on both natural pictures and artificial

images such as psychological patterns. The result indicate

fast and robust saliency detection of our method.

1. Introduction

The first step towards object recognition is object detection. Object detection aims at extracting an object from

its background before recognition. But before performing recognitive feature analysis, how can a machine vision

system extract the salient regions from an unknown background?

Traditional models, by relating particular features with

targets, actually convert this problem to the detection of

specific categories of objects[3]. Since these models are

based on training, the expansibility become the bottleneck

in generalized tasks. Facing unpredictable and innumerable

categories of visual patterns, a general purpose saliency detection system is required. In other words, the saliency detector should be implemented with the least reference on

statistical knowledge of the objects.

How is the saliency detection process achieved in human visual system? It is believed that two stages of visual

processing are involved: first, the parallel, fast, but simple...