Research On Reconstruction Algorithms For Images In Irregular Samplings In Space Remote Sensing

A remote sensing system is often influenced by many factors，and the result is thedegradation of images. The vibrations of the platform and the irregularities in sensorsposition are often the primary causes for image degradation. Satellite images aregenerally acquired by a sweeping bar of sensors known as the time delay integrator（TDI）. This acquisition called push-broom is widely applied in aerospace and earthimaging applications nowadays, because it provides high resolution and high lightutilization efficiency. As a consequence of this progressive acquisition mode, thevibrations of the satellite make two adjacent pixels dislocated and result in perturbedsampling sets. These kind of irregular samplings are responsible of the geometrydistortions of images. Based on the imaging process, image restoration technologyconstructs mathematical model and estimates original image along its inverse directionfor further processing.In this paper, firstly, through analyzing several ways of satellite vibrations whichcause quality degradation of remote sensing images, the irregular sampling problem isexplained and the corresponding mathematical imaging degradation model isconstructed.the model is based on the linear space variant degradation model.In addition, making use of the characteristics of the Toeplitz matrix and cyclicmatrix, a direct reconstruction algorithm for band limited images that considers irregularsampling is proposed, in which the Fast Fourier Transform（FFT）can be used.Furthermore, a1-D irregular to regular sampling algorithm is explored and extended to2-D situation，which is called Adaptive-weights Conjugate gradient on Toeplitz-matrix（ACT） algorithm. This algorithm possesses a quicker convergence; it can also beapplied in degraded images which are noisy, blurred and irregular sampled.Finally, digital simulation experiments focused on the restoration of satell iteimages are carried out to examine the performance of restorations and verify thealgorithm. By employing the degradation model of irregular sampling, noising andblurring, degraded images are acquired. These images are then restored by the twoalgorithms previously proposed. The results of digital simulation experiment prove theeffectiveness of the algorithm and the superiority in the rate of convergence withcomparison to that of the classical conjugate gradient iteration method. After all,evaluation metrics are employed to show the pertinence of the two algorithms.