Detecting Salient Regions in Image Using Visual Attention.

Date of Submission

December 2016

Date of Award

Winter 12-12-2017

Institute Name (Publisher)

Indian Statistical Institute

Document Type

Master's Dissertation

Degree Name

Master of Technology

Subject Name

Computer Science


Electronics and Communication Sciences Unit (ECSU-Kolkata)


Chanda, Bhabatosh (ECSU-Kolkata; ISI)

Abstract (Summary of the Work)

We see a wide range of images everyday. In fact around 1 Giga bits of data enters our eyes every second. However we do not focus on all of this data. Our brain reduces the amount of visual data through high-level cognitive and complex processes. This process is called visual attention and a selection mechanism lies at the core of the process.In the method of salient region detection here, bottom up approach has been used. So no prior information or assumption on the object of interest is used. The algorithm tries to detect objects in images that are different from the rest of the image and consequently capture our attention.In the method here we first pre-process the image by doing albedo correction. We then do a bilateral filtering of the image in Lab space. We find the mean of each of the individual L, a and b channels and then subtract the mean from each of the L, a and b channels. This acts as a band pass filter. We then create the saliency map by taking the L2 norm of the mean subtracted L, a and b channels. Finally we divide the saliency map into superpixels of appropriate size and select the most salient superpixels that are not lying on the boundary and get information regarding the object of attention. The resulting segmented image gives the salient region that is more likely to grab our attention in case if such an image is presented to us.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


This document is currently not available here.