Skip to content
Matlab source code for "Structural Dissimilarity based Saliency model"
MATLAB C++ Java Shell Makefile
Branch: master
Clone or download
Latest commit 7b6b8d5 May 3, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
SDS_EvaluationProject Update readme.md Oct 10, 2018
SDS_demo Bug fixed. Apr 8, 2019
.gitattributes Update .gitattributes Oct 10, 2018
README.md Update README.md May 3, 2019
results.png Add files via upload Mar 21, 2019

README.md

SDS

Source code for the paper "Saliency detection based on structural dissimilarity induced by image quality assessment model", Yang Li and Xuanqin Mou, Journal of Electronic Imaging, 28(2) 023025 (3 April 2019)

SDS means structural-dissimilarity-based saliency.

Yang Li: liyang2012@stu.xjtu.edu.cn

Xuanqin Mou: xqmou@mail.xjtu.edu.cn

The offical website of this code is http://gr.xjtu.edu.cn/web/xqmou/sds_saliency

Citation: @article{ author = { Yang Li,Xuanqin Mou}, title = {Saliency detection based on structural dissimilarity induced by image quality assessment model}, volume = {28}, journal = {Journal of Electronic Imaging}, number = {2}, pages = {023025 - 1 - 14}, year = {2019}, doi = {10.1117/1.JEI.28.2.023025}, }

Results

image

Scores on the MIT300 benchmark: Normalized Scanpath Saliency metric (NSS): 1.3880; AUC (Judd) metric: 0.8148; AUC (Borji) metric: 0.7646; Similarity metric: 0.5183; Cross-correlation metric: 0.5298; shuffled AUC metric: 0.6004; KL metric: 0.8903; Earth Mover Distance metric: 3.1696;

SDS demo

./SDS_demo/demo.m % the demo script showing the usage of the functions SDS_GM.m and SDS_LC.m

./SDS_demo/SDS_GM.m % the SDS model based on the gradient magnitude

./SDS_demo/SDS_LC.m % the SDS model based on the local contrast

Evaluation Project

Usage:

  1. Press F5 and you will get the results of the proposed SDS_GM on the Toronto dataset, i.e., AIM120.

  2. One can add the other models into this project.

    a. Add a new folder: ./MethodCode/YourModel

    b. Add a get-yourmodel-map function in the GetSingleSaliencyMap.m

    c. set the MethodName as "MethodNames = {'YourModel'};"

You can’t perform that action at this time.