A Survey on Visual Saliency Detection and Computational Methods

Abstract

Abstract-In computer vision saliency detection comprises wide range of methods to detect salient object present in the image. These methods focus how important object can be detect from the image. Results of these methods are fully dependent upon the type and quality of input image. In this paper saliency detection methods, Fixation prediction models, saliency map generated through various saliency detection algorithms and its evaluation measures are analyzed.Keyword -Saliency, Seam carving, Saliency Map, Fixation Prediction, Object Detection I. INTRODUCTION With the advancement of computer science and information technology information can be transfer with the use of images. Saliency region detection and object segmentation is also known as salient object detection. To preserve structure of important objects present in the image saliency detection algorithm are used. In image resizing accurate saliency map helpful to find out accurate salient object that must be preserve during image resizing in this way improved algorithm can be developed with accurate result where saliency is required to detect. Accurate saliency detection not only improves performance of the algorithm but also reduces computational time. Methods under saliency detection are classified in to two types of space domain based and frequency domain based [1]. Human visual system also has capability to provide information about salient object in the image. In computer vision saliency detection and salient object segmentation are the process through which most salient object can be detected and boundary of salient object can accurately defined [2]. Literature of saliency detection suggests two models: Top-down model and bottom up model.Bottom up approach pay more attention in color, intensities and edge orientation and top down approach pay more attention on high level features such as face, body and text and their structure. In order to produce better saliency map bottom up approach combines top down approach.Proposed models of saliency detection have been used to measure the exclusiveness of a location for a salient object which mostly attracts human attention [3]. Researchers are trying to improve the efficiency of salient map which can detect large salient region in image, well defined boundaries of object, uniformly highlight full salient object, disregard noise and other blocking artifacts in the image and also should be capable to output full resolution saliency map. Most of the methods in saliency detection are useful for detection of region of interest (ROI). ROI of an image contains pixels which are similar to each other and are capable to fascinate attention of human eyes.In [4], Liu et al. proposed region based pixel-wise saliency detection algorithm which can detect prominent object or region from the image and generate full resolution energy map. Saliency detection models and algorithm suggested by the researchers are widely used in various multimedia disciplines like image compression [5], content aware image resizing [2], image adapt...

Two-stage salient object identification and segmentation based on irregularity

“…In these approaches, it was considered that that salient points usually of higher change in frequency domain both in magnitude and in orientation. For more information about the available methods, many surveys are available such as [19,23,28,29,62], and [24].…”

Dim Target Detection in Infrared Images Using Saliency Algorithms

“…Specifically, research progress of image saliency has been analyzed in [11]. Various methods for saliency detection in color images have been reviewed in [12].…”