Tian Gao

Optical remote sensing imagery is a vital data source for large-scale monitoring in polar regions. However, harsh imaging conditions often result in images with texture degradation. This imagery quality dilemma severely constrains the reliability of downstream analysis. Due to the uniqueness of the polar environment, current enhancement methods tend to introduce artifacts or suffer from limited domain generalization. To address these challenges, we propose a blind perceptual metric-driven enhancement (BPME) method for degraded polar optical imagery. Specifically, we first design a zero-reference metric system to precisely quantify degradation patterns. Subsequently, we construct the metric feature space to derive a blind perceptual metric prior (BPM-P) via clustering-based mapping. Finally, guided by the proposed BPM-P, a selective mechanism is employed to improve texture information adaptively. Comprehensive experiments confirm that the BPME outperforms state-of-the-art algorithms, as indicated by both qualitative and quantitative evaluations. The method also exhibits outstanding performance in downstream tasks such as contour extraction and ice flow velocity tracking. As a physically interpretable framework, BPME provides a robust solution for image degradation in polar remote sensing, effectively enhancing the interpretability and usability of optical remote sensing imagery.

Due to significant nonlinear radiometric differences (NRD) and severe geometric distortions (e.g., translation, scale, and rotation) among multimodal images (e.g., optical, infrared, SAR), image registration remains a challenging problem. This paper proposes a robust matching method based on self-similarity adjacent self-convolution features (SASF), consisting of two key steps. In the first stage, a uniform distribution of feature points in multimodal images is extracted using a FAST feature extraction operator based on normalized filtering, and geometric correction is performed using prior reference geographic location information of remote sensing images to achieve coarse registration. In the second stage, addressing the significant NRD, a novel structural feature SASF is constructed. Based on this, a rapid structural similarity measurement method ($Fast-SSIM_{SASF}$) is proposed, which utilizes 3D fast Fourier transform techniques to enhance image matching efficiency. The proposed SASF algorithm is evaluated for performance on multiple sets of multimodal remote sensing images. Experimental results demonstrate that this method outperforms many existing excellent algorithms in terms of matching effectiveness. Finally, the SASF algorithm is tested on four pairs of entire-scene multimodal remote sensing satellite images, demonstrating its robustness to nonlinear radiometric differences and geometric distortions.

The current traditional image matching methods are difficult to adapt to multimodal image matching, in which the main influencing factors are the nonlinear radiation difference (NRD) and geometric deformation between different sensors. To address this challenge, this article proposes a multimodal matching method based on a local window filter that sharpens the edge fusion phase. The steps are listed below: (1) Enhance the edge information of the image using an improved local window filter, construct the scale space of edge information enhancement, and then calculate the maximum and minimum moments of the phase congruency of the enhanced image, which constitutes the fusion moment of edge enhancement. (2) Introduce sharpness features to enhance the edge structure and then weighted fusion with the edge enhancement fusion moment calculated in (1) to generate sharpened fusion phase features. (3) Combine the optimized log-polar coordinates for description, adopt the Euclidean distance as the similarity measure for matching, and finally use the fast sample consistency algorithm for mismatch rejection. This study uses a collection of multimodal remote sensing images showing NRD and geometric deformation, with comparative experiments conducted using LNIFT, OSS, OS-SIFT RIFT and CMM-Net methods. The experimental results show that the method proposed in the article outperforms other similar algorithms in terms of the number and precision of matching points. The number of correctly matched points is improved by more than 2 times on average, and the root mean square error (RMSE) reaches average 1.24 pixels, which proves the algorithm has better robustness and accuracy.