Feature matching between image pairs is a fundamental problem in computer vision that drives many applications, such as SLAM. Recently, semi-dense matching approaches have achieved substantial performance enhancements and established a widely-accepted coarse-to-fine paradigm. However, the majority of existing methods focus on improving coarse feature representation rather than the fine-matching module. Prior fine-matching techniques, which rely on point-to-patch matching probability expectation or direct regression, often lack precision and do not guarantee the continuity of feature points across sequential images. To address this limitation, this paper concentrates on enhancing the fine-matching module in the semi-dense matching framework. We employ a lightweight and efficient homography estimation network to generate the perspective mapping between patches obtained from coarse matching. This patch-to-patch approach achieves the overall alignment of two patches, resulting in a higher sub-pixel accuracy by incorporating additional constraints. By leveraging the homography estimation between patches, we can achieve a dense matching result with low computational cost. Extensive experiments demonstrate that our method achieves higher accuracy compared to previous semi-dense matchers. Meanwhile, our dense matching results exhibit similar end-point-error accuracy compared to previous dense matchers while maintaining semi-dense efficiency.

Semi-dense feature matching methods have shown strong performance in challenging scenarios. However, the existing pipeline relies on a global search across the entire feature map to establish coarse matches, limiting further improvements in accuracy and efficiency. Motivated by this limitation, we propose a novel pipeline, CasP, which leverages cascaded correspondence priors for guidance. Specifically, the matching stage is decomposed into two progressive phases, bridged by a region-based selective cross-attention mechanism designed to enhance feature discriminability. In the second phase, one-to-one matches are determined by restricting the search range to the one-to-many prior areas identified in the first phase. Additionally, this pipeline benefits from incorporating high-level features, which helps reduce the computational costs of low-level feature extraction. The acceleration gains of CasP increase with higher resolution, and our lite model achieves a speedup of $\sim2.2\times$ at a resolution of 1152 compared to the most efficient method, ELoFTR. Furthermore, extensive experiments demonstrate its superiority in geometric estimation, particularly with impressive cross-domain generalization. These advantages highlight its potential for latency-sensitive and high-robustness applications, such as SLAM and UAV systems. Code is available at https://github.com/pq-chen/CasP.

The multi-modal remote sensing foundation model (MM-RSFM) has significantly advanced various Earth observation tasks, such as urban planning, environmental monitoring, and natural disaster management. However, most existing approaches generally require the training of separate backbone networks for each data modality, leading to redundancy and inefficient parameter utilization. Moreover, prevalent pre-training methods typically apply self-supervised learning (SSL) techniques from natural images without adequately accommodating the characteristics of remote sensing (RS) images, such as the complicated semantic distribution within a single RS image. In this work, we present SkySense V2, a unified MM-RSFM that employs a single transformer backbone to handle multiple modalities. This backbone is pre-trained with a novel SSL strategy tailored to the distinct traits of RS data. In particular, SkySense V2 incorporates an innovative adaptive patch merging module and learnable modality prompt tokens to address challenges related to varying resolutions and limited feature diversity across modalities. In additional, we incorporate the mixture of experts (MoE) module to further enhance the performance of the foundation model. SkySense V2 demonstrates impressive generalization abilities through an extensive evaluation involving 16 datasets over 7 tasks, outperforming SkySense by an average of 1.8 points.

Prior studies on Remote Sensing Foundation Model (RSFM) reveal immense potential towards a generic model for Earth Observation. Nevertheless, these works primar-ily focus on a single modality without temporal and geo-context modeling, hampering their capabilities for diverse tasks. In this study, we present SkySense, a generic billion-scale model, pretrained on a curated multimodal Remote Sensing Imagery (RSI) dataset with 21.5 million temporal sequences. SkySense incorporates a factorized multimodal spatiotemporal encoder taking temporal sequences of opti-cal and Synthetic Aperture Radar (SAR) data as input. This encoder is pretrained by our proposed Multi-Granularity Contrastive Learning to learn representations across different modal and spatial granularities. To further enhance the RSI representations by the geo-context clue, we introduce Geo-Context Prototype Learning to learn region-aware prototypes upon RSI's multimodal spatiotemporal features. To our best knowledge, SkySense is the largest Multi-Modal RSFM to date, whose modules can be flexibly combined or used individually to accommodate various tasks. It demonstrates remarkable generalization capabilities on a thor-ough evaluation encompassing 16 datasets over 7 tasks, from single- to multimodal, static to temporal, and classification to localization. SkySense surpasses 18 recent RSFMs in all test scenarios. Specifically, it outperforms the latest models such as GFM, SatLas and Scale-MAE by a large margin, i.e., 2.76%, 3.67% and 3.61% on average respectively. We will release the pretrained weights to facilitate future research and Earth Observation applications.

Long-tailed recognition (LTR) aims to learn balanced models from extremely unbalanced training data. Fine-tuning pretrained foundation models has recently emerged as a promising research direction for LTR. However, we observe that the fine-tuning process tends to degrade the intrinsic representation capability of pretrained models and lead to model bias towards certain classes, thereby hindering the overall recognition performance. To unleash the intrinsic representation capability of pretrained foundation models, in this work, we propose a new Parameter-Efficient Complementary Expert Learning (PECEL) for LTR. Specifically, PECEL consists of multiple experts, where individual experts are trained via Parameter-Efficient Fine-Tuning (PEFT) and encouraged to learn different expertise on complementary sub-categories via the proposed sample-aware logit adjustment loss. By aggregating the predictions of different experts, PECEL effectively achieves a balanced performance on long-tailed classes. Nevertheless, learning multiple experts generally introduces extra trainable parameters. To ensure parameter efficiency, we further propose a parameter sharing strategy which decomposes and shares the parameters in each expert. Extensive experiments on 4 LTR benchmarks show that the proposed PECEL can effectively learn multiple complementary experts without increasing the trainable parameters and achieve new state-of-the-art performance.

Selection of a reference image is an important step in color balancing. However, the past research and currently available methods do not focus on it, leading to the lack of an effective way to select the reference image for color balancing in remote sensing imagery mosaic. This letter proposes a novel automatic reference image selection method that aims to select the reference images by assessing multifactors according to the land surface types of the target images. The proposed method addresses the limitations caused by the use of a single assessment factor as well as the selection of a single image as the reference in traditional methods. In addition, the proposed method has a wider range of applications than those requiring no reference image. The visual experimental results indicate that the proposed method can select the suitable reference images, which benefits the color balancing result, and outperforms the other comparative methods. Moreover, the absolute mean value of skewness metric of the proposed method is 0.0831, which is lower than the values of the other comparison methods. It indicates that the result of the proposed method had the best performance in the color information. The quantitative analyses with the metric of absolute difference of mean value indicate that the proposed method has a good ability in maintaining the spectral information, and the spectral changing rates had been reduced at least 10.66% by the proposed method when compared with the other methods.

A new mixed radiometric normalization (MRN) method is introduced in this paper which aims to eliminate the radiometric difference in image mosaicking. The radiometric normalization methods can be classified as the absolute and relative approaches in traditional solutions. Though the absolute methods could get the precise surface reflectance values of the images, rigorous conditions required for them are usually difficult to obtain, which makes the absolute methods impractical in many cases. The relative methods, which are simple and practicable, are more widely applied. However, the standard for designating the reference image needed for these methods is not unified. Moreover, the color error propagation and the two-body problems are common obstacles for the relative methods. The proposed MRN approach combines absolute and relative radiometric normalization methods, by which the advantages of both can be fully used and the limitations can be effectively avoided. First, suitable image after absolute radiometric calibration is selected as the reference image. Then, the invariant feature probability between the pixels of the target image and that of the reference image is obtained. Afterward, an adaptive local approach is adopted to obtain a suitable linear regression model for each block. Finally, a bilinear interpolation method is employed to obtain the radiometric calibration parameters for each pixel. Moreover, the CIELAB color space is adopted to evaluate the results quantitatively. Experimental results of ZY-3, GF-1, and GF-2 data indicate that the proposed method can eliminate the radiometric differences between images from the same or even different sensors.

Most sun-synchronous satellites acquire images when the illumination source is located in the southeast direction, which is contrary to human visual habits. Therefore, when observing such images, the false topographic perception phenomenon (FTPP) occurs, particularly in large-relief regions. FTPP should be corrected because it may induce misunderstandings of terrain features, thus affecting related decision making. Unlike most existing algorithms, an automatic algorithm focusing on altering the illumination direction is proposed on the basis of topographic correction. The purpose of the algorithm is to alter the illumination direction from SE to northwest. The original and FTPP-corrected images can be quantitatively connected by establishing the relationship between these two images after topographic correction by using the C model such that the formula for calculating the DN (Digital Number) value of the FTPP-corrected image can be derived. An online survey of 323 volunteers using two sets of test data demonstrates that the success rate of the topographic-correction-based approach is as high as 85%. In comparative trials, the proposed approach displays a relatively strong ability to preserve the spatial details and the spectral characteristics of the original image.