Estimation of Local and Global Superpixel Covariance for Salient Object Detection in Low Contrast Images

Abstract

Salient object detection has become a hot topic in computer vision as it can substantially facilitate a wide range of applications. Conventional salient object detection models primarily rely on low-level image features, which may face great difficulties in low lighting scenarios. This paper proposes to estimate the saliency of low contrast images via covariance features. The input image is firstly decomposed into superpixel regions to estimate their covariances. Then, the local and global image saliency can be calculated using the covariance features respectively. Finally, a graph-based diffusion process is performed to refine the saliency maps. Extensive experiments have been conducted to evaluate the performance of the proposed model against eleven state-of-the-art models on five benchmark datasets and a nighttime image dataset.