Salient Object Detection from the Video Streams Using an Enhanced Water Flow Approach

Abstract

In this paper, we formulate the issue in the salient object detection (SOD) on the video stream using a novel proposed method named," Water Flow Driven using Minimum Barrier Distance". By using this method the two objective are been state which are been rectified in our work. The major part relies on finding the larger area in the frame which is used for false positive for the fixation prediction on the image. When the salient object are located near the boundary region, the accuracy are lacking, in order to increases the accuracy our work performance in the better improvement in the planned work. And the second objective is the RGB resolution here the foreground and the background are having the similar color, and then it is difficult to detect the image salient these issues are facing to the degradable in the image post processing of the salient image. The final experimental study and the performance metrics state the Precision-Recall, Mean Absolute and the F-measure are been improved when the comparison of the existing work. The result validate the proposed work is having the PASCAL-S dataset for the image feature extraction and there are very robust to the resolution.