DETECTING SATURATED PIXELS AND BLOOMING EFFECTS FROM SATELLITE IMAGES BASED ON IMAGE STATISTICS AND SPATIAL IMAGE QUALITY ANALYSIS

초록

Identifying saturated and blooming pixels within optical satellite imagery can help users with image interpretation and further processing. Saturation refers to masking pixels that exceed the representable brightness range. This occurs due to sensor calibration issues, high reflectance, shooting altitude, and other factors. Blooming refers to the phenomenon in which an over-charged signal spills over to nearby pixels when saturation occurs in a Charge-Coupled Device (CCD) sensor. Saturation and blooming in satellite imagery can lead to distortion in visibility and pixel information loss. For this reason, earth observation satellite image vendors, such as Planet, provide users with a saturation mask within Unusable Data Mask (UDM). However, this data does not simultaneously detect the presence of blooming objects. In this study, we propose a method for detecting saturation and blooming in medium-resolution satellite imagery based on image statistics and edge response. We used two Landsat-8 OLI Images, which provided a location of saturated pixels in Quality Assessment (QA) mask data. For Saturated pixel detection, we considered QA mask as ground truth and compared our detection results quantitatively and qualitatively. For blooming effect detection, we choose blooming areas manually and evaluate our results. In conclusion, saturation mask exhibited an average accuracy of 84% when compared to the QA mask. In the manually selected blooming regions, our method demonstrated a detection accuracy of 100%. © 2023 ACRS. All Rights Reserved.

키워드