An efficient method to detect defective detectors of SPOT 4 SWIR band using iterative high-pass filtering

An efficient method to detect defective detectors of SPOT 4 SWIR band is presented. About one hundred of SWIR detectors out of 3000 SPOT 4 detectors malfunction, which degrade the quality of image. It is often difficult for conventional statistical method to effectively detect malfunctioning detectors because of heterogeneous nature of the earth surface. We propose a new method for detection of the defective detectors by flattening the baseline through high-pass filtering in the wave number domain. Iteration is necessary to improve performance. Three SPOT 4 images with different sensors were used to evaluate the performance of the proposed method. A total of 338 defective detectors out of 352 actual defective detectors in the three images were successfully detected. This result showed a significant improvement of the detection capability with an overall detection accuracy of 96%. This approach was less effective in the regions of waters and clouds. Omission error of the proposed method was more serious than commission error. While 14 defective detectors were not detected, 4 detectors were wrongly detected. These errors did not significantly affect the overall quality of corrected image.