Saliency Detection Model for Low Contrast Images Based on Amplitude Spectrum Analysis and Superpixel Segmentation

Traditional saliency detection models face great challenges towards low contrast images with low signal-to-noise ratio property. In this circumstance, it is difficult to extract effective visual features to describe salient information in image. This paper proposes a saliency detection model for low contrast images utilizing efficient features both from frequency domain and spatial domain. The input image is firstly transformed into frequency domain to calculate the amplitude spectrum by a median filter, aiming to suppress the information from non-salient regions. Then, a superpixel based feature extraction method is utilized to generate saliency map via both local and global spatial information. Experiments are carried on the low contrast image dataset to demonstrate the effectiveness of the proposed saliency detection model over other eight state-of-the-art saliency models.