Automatic Detection of Blurred Areas for Remote Sensing Imagery.
Cheng Su,Zeyu Xu,Fan Shen,Xiaocan Zhang
DOI: https://doi.org/10.1109/lgrs.2023.3324034
IF: 5.343
2023-01-01
IEEE Geoscience and Remote Sensing Letters
Abstract:In the interest of national security, certain geographical targets, such as military installations, nuclear facilities, and secret agencies, must be concealed in satellite images prior to public release. This involves the modification of specific regions within the satellite image through techniques, such as pixelation, replacement, and blurring to obscure the details of sensitive targets. However, the technical process may not completely eliminate all traces of the sensitive information, potentially enabling a satellite imagery analyst to identify the concealed targets. This letter presents an automatic detection method for identifying blurred areas, which may correspond to sensitive targets, within remote sensing images. The experimental results demonstrate that this method effectively detects blurred areas with minimal omissions. Furthermore, this letter explores the relationship between the level of visual blurring and the difficulty of automatic detection, emphasizing the importance of selecting an appropriate blur level (BL) for images with varying degrees of heterogeneity.