An improved active contour model for salient object detection using edge cues

Abstract

In this paper, we solve the problem of salient object detection by using an ensemble mechanism of edge detection and active contours. Edge cues are used to provide a solution to the initialisation problem of active contour. The active contour method suffers from the initialisation problem. If the initial contour lies in a region with low probability of salient object, the final salient object detection provides inaccurate results. In this paper, the problem is addressed by generating a binary mask using Sobel edge detection method which acts as the initial contour. The binary mask makes sure that the contour lies in the region with high probability of finding a salient object. This work is an improvement upon the active contour model. The method is simple and fast and follows basic human intuition to find salient objects. The proposed work is compared against seven recent works and gives better results in terms of precision, recall and false positive rates. © Springer Nature Singapore Pte Ltd. 2019.