Salient object detection using task simulation as a new input

Abstract

Saliency or the salient region changes in the human vision system depending on the type of its behavior and task. That is, the salient region in human vision system is not independent, but dependent on other parameters. If a saliency detection algorithm intends to work like the human vision system, it must have an input as its vision in order to detect that salient region or salient object according to that input. The proposed algorithm of this article (Salient Object Detection using Task Simulation based on Angle) is indeed the updated version of SOD-TS algorithm. In this method we have tried to simulate the task as an angle parameter in order to be able to detect the object independent of its size and rotation. In our proposed method, the algorithm detects the most salient object with regard to the applied task. This method can be used in detecting salient objects, detecting different types of ships, and different types of airplanes, and in edge detection. One of the most important advantages of this approach is its very high speed.