An Adaptive Thresholding Method with Local Binary Patterns Features for Image Reconstruction

Remote sensing images can experience degradation in image quality due to various factors, both inherent to the imaging process and external to it. In general, image restoration can help improve image quality. The Maximum a Posteriori (MAP) algorithm integrates a priori knowledge and observational data during the image reconstruction process, enhancing the clarity of the resulting image while maintaining its authenticity. In this paper, an Adaptive-thresholding MAP (At-MAP) algorithm is presented, which combines adaptive thresholding based on Huber-function with local binary pattern (LBP) features for remote sensing image reconstruction. Being guided by the magnitudes of brightness (DN values), threshold values for the Huber regularization term are adaptively selected. The experiments in this study include both simulated datasets and actual remote sensing images. The results demonstrate that the proposed method outperforms other state of the art methods both in subjective visual comparison and in objective numerical comparison. It outperforms the deep learning algorithm by about 5% in SNR and by about 25% in blur matrix. The proposed method holds significant potential applications as it enhances spatial resolution, mitigates ringing artifact, and preserves image fidelity.