Search

Speckle suppression and detail preservation are a unity of contradiction during synthetic aperture radar (SAR) despeckling. In this article, an effective despeckling method is proposed based on a self-adaptive neighbourhood, which takes into account the trade-off between homogeneous region suppression and point target preservation. The size and shape of the neighbourhood depend on an adaptive threshold that is treated as a linear function of a saliency map and a threshold range estimated by the Monte Carlo algorithm based on a given significance level and the number of looks in SAR. If the saliency of the current pixel is large, which always means that the current pixels should belong to a point target, a linear object, or an edge, a relatively large threshold will be assigned to search a neighbourhood with a small size. Conversely, pixels with a small saliency value will obtain a non-local neighbourhood with a large size. A maximum-likelihood estimation is employed to estimate the real radar reflectivity based on the pixels in the adaptive neighbourhood. The comparison experiments on simulated and actual SAR data validate the effectiveness of the proposed algorithm and demonstrate its overall speckle filtering characteristics compared with other algorithms. The visual and numerical experimental results show that the proposed despeckling method provides superior performance compared with several state-of-the-art despeckling methods. [ABSTRACT FROM PUBLISHER]

Hyperspectral Unmixing (HU), which involves decomposing mixed pixels in hyperspectral images into endmembers and abundance fractions, plays a critical role in spectral analysis. Recent advancements in deep learning methods have shown successful applications in hyperspectral analysis. This paper proposes an end-to-end network structure that combines a Convolutional Neural Network (CNN) and a Multilayer Perceptron (MLP). The 3D-CNN extracts high-level features from both spatial and spectral neighbourhoods of hyperspectral image patches, while the MLP maps these abstracted features to abundance fractions. Considering the challenge of acquiring manual labels, we adopt a Semi-Supervised learning approach that leverages unlabelled data to enhance training. We divide our training data into labelled and unlabelled subsets, defining distinct loss functions for each. Additionally, we introduce a neighbourhood similarity constraint to fully exploit spatial context information. To address scenarios where no labelled data is available, we employ Generative Adversarial Networks (GANs) to generate synthetic labelled samples. Experimental evaluations conducted on both synthetic and real datasets validate the effectiveness of our proposed method, which outperforms several commonly used and state-of-the-art HU methods. [ABSTRACT FROM AUTHOR]