Papers › A model of saliency-based visual attention for rapid scene analysis

A model of saliency-based visual attention for rapid scene analysis

1 Nov 1998IEEE Transactions on Pattern Analysis and Machine Intelligence 1998 11archive 2025-07-28

L. Itti, C. Koch, E. Niebur

A visual attention system, inspired by the behavior and the neuronal architecture of the early primate visual system, is presented. Multiscale image features are combined into a single topographical saliency map. A dynamical neural network then selects attended locations in order of decreasing saliency. The system breaks down the complex problem of scene understanding by rapidly selecting, in a computationally efficient manner, conspicuous locations to be analyzed in detail.

PaperPDF

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Saliency PredictionScene UnderstandingVideo Saliency Detection

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Video Saliency Detection MSU Video Saliency Prediction ITTI AUC-J 0.811 #14 of 14 Archive leaderboard report
Video Saliency Detection MSU Video Saliency Prediction ITTI CC 0.572 #14 of 14 Archive leaderboard report
Video Saliency Detection MSU Video Saliency Prediction ITTI FPS 2.23 #14 of 14 Archive leaderboard report
Video Saliency Detection MSU Video Saliency Prediction ITTI KLDiv 0.799 #14 of 14 Archive leaderboard report
Video Saliency Detection MSU Video Saliency Prediction ITTI NSS 1.34 #14 of 14 Archive leaderboard report
Video Saliency Detection MSU Video Saliency Prediction ITTI SIM 0.544 #14 of 14 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections