Abstract

The extraction of a region of interest is an important component of remote sensing image analyses. Driven by practical applications, a good region of interest (ROI) needs to have three properties: uniformly highlighting entire ROIs, well-defined boundaries, and good stability against noisy data. Motivated by these requirements, we propose a ROI extraction model based on saliency analysis of co-occurrence histogram (SACH) in high spatial resolution remote sensing images. First, a co-occurrence histogram is utilized to capture the global and local distribution of intensity values. Secondly, our model estimates the saliency of the co-occurrence histogram by utilizing a logarithm function. Thirdly, a saliency-enhanced method based on moving K-means aggregation is utilized to establish well-defined boundary for ROIs and improve immunity to noise. Finally, ROIs are segmented from the saliency maps of original images, which are acquired from the saliency of the co-occurrence histogram. In the experimental part, we compare our model with nine other extraction models by applying the models to clean images and to images corrupted by noises. The experimental results show that compared to the nine competing models, SACH model better defines the boundaries of target ROIs and gets more entire ROIs. Furthermore, SACH model is also robust against images corrupted by Gaussian and Salt and Pepper noises.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call