Defective CCDs detection and image restoration based on inter-band radiance interpolation for hyperspectral imager

A 2D detector array is used popularly to acquire image in spatial and spectral dimension simultaneously for hyperspectral imager. The detector array will be malfunctioned gradually and partially after long-term operations. These defective CCDs will cause vertical stripes in images. But it's not cost effective to replace the detector due to a few of defects. In this article, we propose an algorithm including two parts for hyperspectral image restoration. One is the CCDs defect parts detection according to their radiance deviation, and another is the image restoration based on inter-band radiance interpolation using Lagrange polynomial. The detection process of finding CCDs defect parts for an imager must be conducted periodically to update the CCDs health status. HOPE images with simulated defective CCDs of various performance decay level are applied for validation. We found the accuracy for images with homogeneous ground feature is higher than the ones with non-homogeneous feature. And defect CCDs with performance decay of 10% still can be designated precisely. Restoration accuracy of pixel radiance is presented for various spectral bands using proposed algorithm. We also perform the image reconstruction using interpolation of spatial neighboring pixels. Radiance deviation for restored pixels is compared between both methods. Proposed algorithm can handle the images taken by hyperspectral imager with adjoining defective CCDs both in spatial and spectral. However, the method using interpolation of neighboring pixels can't. Applying the purposed algorithm on hyperspectral images, the imager can continue operating like a good one though there are a few of defects in detector.