arXiv subjects

Shihong Wang

Publications and source records attributed to Shihong Wang.

16 recordsLinked to original sources

arXiv2026

FIRM: Fine-Grained Intra-Token Representation of Masks for Remote Sensing Reasoning Segmentation

Reasoning segmentation requires multimodal large language models (MLLMs) to translate implicit instructions into precise pixel-level masks. MLLMs encode an image as visual tokens, each of which merges a group of image patches. In remote sensing images, small targets, thin structures, and adjacent instances can occupy different parts of the same visual token. Assigning a single binary mask label to such a token loses its internal spatial structure, causing nearby targets to merge and object boundaries to become coarse. To bridge this representational gap, we introduce FIRM, a Fine-grained Intra-token Representation of Masks. For each visual token, FIRM predicts a mask code that specifies an $r\times r$ binary sub-cell pattern rather than a single foreground/background label. Given a target identified by the MLLM, the complete grid of mask codes is predicted in one mask pass. Fixed lookup converts the predicted codes into a discrete sub-cell mask, while marginalizing the code distribution yields a soft structural field. To further recover fine-grained boundaries within each sub-cell, we introduce a lightweight continuous renderer that refines this field using pre-merge visual features and image details. Across five reasoning and referring segmentation benchmarks on satellite and UAV images, FIRM achieves leading results, including $70.5/80.5$ gIoU/cIoU on LaSeRS and a $3.0$-point average gain on EarthReason. These results demonstrate the value of explicitly representing intra-token mask patterns for fine-grained MLLM segmentation.

cs.CV↗

arXiv2025

Annotation-Free Open-Vocabulary Segmentation for Remote-Sensing Images

Semantic segmentation of remote sensing (RS) images is pivotal for comprehensive Earth observation, but the demand for interpreting new object categories, coupled with the high expense of manual annotation, poses significant challenges. Although open-vocabulary semantic segmentation (OVSS) offers a promising solution, existing frameworks designed for natural images are insufficient for the unique complexities of RS data. They struggle with vast scale variations and fine-grained details, and their adaptation often relies on extensive, costly annotations. To address this critical gap, this paper introduces SegEarth-OV, the first framework for annotation-free open-vocabulary segmentation of RS images. Specifically, we propose SimFeatUp, a universal upsampler that robustly restores high-resolution spatial details from coarse features, correcting distorted target shapes without any task-specific post-training. We also present a simple yet effective Global Bias Alleviation operation to subtract the inherent global context from patch features, significantly enhancing local semantic fidelity. These components empower SegEarth-OV to effectively harness the rich semantics of pre-trained VLMs, making OVSS possible in optical RS contexts. Furthermore, to extend the framework's universality to other challenging RS modalities like SAR images, where large-scale VLMs are unavailable and expensive to create, we introduce AlignEarth, which is a distillation-based strategy and can efficiently transfer semantic knowledge from an optical VLM encoder to an SAR encoder, bypassing the need to build SAR foundation models from scratch and enabling universal OVSS across diverse sensor types. Extensive experiments on both optical and SAR datasets validate that SegEarth-OV can achieve dramatic improvements over the SOTA methods, establishing a robust foundation for annotation-free and open-world Earth observation.

cs.CV↗

arXiv2024

Class Similarity Transition: Decoupling Class Similarities and Imbalance from Generalized Few-shot Segmentation

In Generalized Few-shot Segmentation (GFSS), a model is trained with a large corpus of base class samples and then adapted on limited samples of novel classes. This paper focuses on the relevance between base and novel classes, and improves GFSS in two aspects: 1) mining the similarity between base and novel classes to promote the learning of novel classes, and 2) mitigating the class imbalance issue caused by the volume difference between the support set and the training set. Specifically, we first propose a similarity transition matrix to guide the learning of novel classes with base class knowledge. Then, we leverage the Label-Distribution-Aware Margin (LDAM) loss and Transductive Inference to the GFSS task to address the problem of class imbalance as well as overfitting the support set. In addition, by extending the probability transition matrix, the proposed method can mitigate the catastrophic forgetting of base classes when learning novel classes. With a simple training phase, our proposed method can be applied to any segmentation network trained on base classes. We validated our methods on the adapted version of OpenEarthMap. Compared to existing GFSS baselines, our method excels them all from 3% to 7% and ranks second in the OpenEarthMap Land Cover Mapping Few-Shot Challenge at the completion of this paper. Code: https://github.com/earth-insights/ClassTrans

cs.CV↗

arXiv2022

E-LMC: Extended Linear Model of Coregionalization for Spatial Field Prediction

Physical simulations based on partial differential equations typically generate spatial fields results, which are utilized to calculate specific properties of a system for engineering design and optimization. Due to the intensive computational burden of the simulations, a surrogate model mapping the low-dimensional inputs to the spatial fields are commonly built based on a relatively small dataset. To resolve the challenge of predicting the whole spatial field, the popular linear model of coregionalization (LMC) can disentangle complicated correlations within the high-dimensional spatial field outputs and deliver accurate predictions. However, LMC fails if the spatial field cannot be well approximated by a linear combination of base functions with latent processes. In this paper, we present the Extended Linear Model of Coregionalization (E-LMC) by introducing an invertible neural network to linearize the highly complex and nonlinear spatial fields so that the LMC can easily generalize to nonlinear problems while preserving the traceability and scalability. Several real-world applications demonstrate that E-LMC can exploit spatial correlations effectively, showing a maximum improvement of about 40% over the original LMC and outperforming the other state-of-the-art spatial field models.

cs.LG↗

arXiv2019

Security analysis of a self-embedding fragile image watermark scheme

Recently, a self-embedding fragile watermark scheme based on reference-bits interleaving and adaptive selection of embedding mode was proposed. Reference bits are derived from the scrambled MSB bits of a cover image, and then are combined with authentication bits to form the watermark bits for LSB embedding. We find this algorithm has a feature of block independence of embedding watermark such that it is vulnerable to a collage attack. In addition, because the generation of authentication bits via hash function operations is not related to secret keys, we analyze this algorithm by a multiple stego-image attack. We find that the cost of obtaining all the permutation relations of $l\cdot b^2$ watermark bits of each block (i.e., equivalent permutation keys) is about $(l\cdot b^2)!$ for the embedding mode $(m, l)$, where $m$ MSB layers of a cover image are used for generating reference bits and $l$ LSB layers for embedding watermark, and $b\times b$ is the size of image block. The simulation results and the statistical results demonstrate our analysis is effective.

cs.CR↗

arXiv2018

Detecting Directed Interactions of Networks by Random Variable Resetting

We propose a novel method of detecting directed interactions of a general dynamic network from measured data. By repeating random state variable resetting of a target node and appropriately averaging over the measurable data, the pairwise coupling function between the target and the response nodes can be inferred. This method is applicable to a wide class of networks with nonlinear dynamics, hidden variables and strong noise. The numerical results have fully verified the validity of the theoretical derivation.

physics.data-an↗

arXiv2017

Reconstructing nonlinear networks subject to fast-varying noises by using linearization with expanded variables

Reconstructing noise-driven nonlinear networks from time series of output variables is a challenging problem, which turns to be very difficult when nonlinearity of dynamics, strong noise impacts and low measurement frequencies jointly affect. In this Letter, we propose a general method that introduces a number of nonlinear terms of the measurable variables as artificial and new variables, and uses the expanded variables to linearize nonlinear differential equations. Moreover, we use two-time correlations to decompose effects of system dynamics and noise driving. With these transformations, reconstructing nonlinear dynamics of the original system is equivalent to solving linear dynamics of the expanded system at the least squares approximations. We can well reconstruct nonlinear networks, including all dynamic nonlinearities, network links, and noise statistical characteristics, as sampling frequency is rather low. Numerical results fully verify the validity of theoretical derivations.

cond-mat.stat-mech↗

arXiv2017

A secure blind watermarking scheme based on DCT domain of the scrambled image

This paper investigates a secure blind watermarking scheme. The main idea of the scheme not only protects the watermark information but also the embedding positions. To achieve a higher level of security, we propose a sub key generation mechanism based on the singular value decomposition and hash function, where sub keys depend on both the main key and the feature codes of the original image. The different sub keys ensure that the embedding positions are randomly selected for different original images. Watermark is embedded in the Discrete Cosine Transform (DCT) coefficients of the scrambled original image. Simulation results show that such embedded method resolves well the contradiction of imperceptibility and robustness. Based on good correlation properties of chaotic sequences, we design a detection method, which can accurately compute geometric transformation (rotation and translation transformations) parameters. The security analysis, including key space analysis, key sensitivity analysis, cryptanalysis, and the comparison results demonstrate that the proposed watermarking scheme also achieves high security.

cs.CR↗

arXiv2016

Depict noise-driven nonlinear dynamic networks from output data by using high-order correlations

Many practical systems can be described by dynamic networks, for which modern technique can measure their output signals, and accumulate extremely rich data. Nevertheless, the network structures producing these data are often deeply hidden in these data. Depicting network structures by analysing the available data, i.e., the inverse problems turns to be of great significant. On one hand, dynamics are often driven by various unknown facts, called noises. On the other hand, network structures of practical systems are commonly nonlinear, and different nonlinearities can provide rich dynamic features and meaningful functions of realistic networks. So far, no method, both theoretically or numerically, has been found to systematically treat the both difficulties together. Here we propose to use high-order correlation computations (HOCC) to treat nonlinear dynamics; use two-time correlations to treat noise effects; and use suitable basis and correlator vectors to unifiedly depict all dynamic nonlinearities and topological interaction links and noise statistical structures. All the above theoretical frameworks are constructed in a closed form and numerical simulations fully verify the validity of theoretical predictions.

cond-mat.stat-mech↗

arXiv2008

A self-synchronizing stream cipher based on chaotic coupled maps

A revised self-synchronizing stream cipher based on chaotic coupled maps is proposed. This system adds input and output functions aim to strengthen its security. The system performs basic floating-point analytical computation on real numbers, incorporating auxiliarily with algebraic operations on integer numbers.

nlin.CD↗

arXiv2005

A new self-synchronizing stream cipher

A new self-synchronizing stream cipher (SSSC) is proposed based on one-way and nearest neighbor coupled integer maps. Some ideas of spatiotemporal chaos synchronization and chaotic cryptography are applied in this new SSSC system. Several principles of constructing optimal SSSC are discussed, and the methods realizing these principles are specified. This SSSC is compared with several SSSC systems in security by applying chosen-ciphertext attacks. It is shown that our new system can provide SSSC with high security and fairly fast performance.

nlin.CD↗

arXiv2004

Experimental realization of a highly secure chaos communication under strong channel noise

A one-way coupled spatiotemporally chaotic map lattice is used to contruct cryptosystem. With the combinatorial applications of both chaotic computations and conventional algebraic operations, our system has optimal cryptographic properties much better than the separative applications of known chaotic and conventional methods. We have realized experiments to pratice duplex voice secure communications in realistic Wired Public Switched Telephone Network by applying our chaotic system and the system of Advanced Encryption Standard (AES), respectively, for cryptography. Our system can work stably against strong channel noise when AES fails to work.

nlin.CD↗

arXiv2004

Periodicity of chaotic trajectories in realizations of finite computer precisions and its implication in chaos communications

Fundamental problems of periodicity and transient process to periodicity of chaotic trajectories in computer realization with finite computation precision is investigated by taking single and coupled Logistic maps as examples. Empirical power law relations of the period and transient iterations with the computation precisions and the sizes of coupled systems are obtained. For each computation we always find, by randomly choosing initial conditions, a single dominant periodic trajectory which is realized with major portion of probability. These understandings are useful for possible applications of chaos, e.g., chaotic cryptography in secure communication.

nlin.CD↗

arXiv2003

Chaos-based cryptograph incorporated with S-box algebraic operation

Error function analysis is an effective attack against chaotic cryptograph [PRE 66, 065202(R) (2002)]. The basin structure of the error function is crucial for determining the security of chaotic cryptosystems. In the present paper the basin behavior of the system used in [21] is analyzed in relation with the estimation of its practical security. A S-box algebraic operation is included in the chaotic cryptosystem, which considerably shrinks the basin of the error function and thus greatly enhances the practical security of the system with a little computational expense. PACS numbers: 05.45.Vx, 05.45.Ra

nlin.CD↗

arXiv2003

A spatiotemporal-chaos-based cryptosystem taking advantages of both synchronous and self-synchronizing schemes

Two-dimensional one-way coupled map lattices are used for cryptograph where multiple space units produce chaotic outputs in parallel. One of the outputs plays the role of driving for synchronization of the decryption system while the others perform the function of information encoding. With this separation of functions the receiver can establish a self-checking and self-correction mechanism, and enjoys the advantages of both synchronous and self-synchronizing schemes. A comparison between the present system with the system of Advanced Encryption Standard, AES, is presented in the aspect of channel noise influence. Numerical investigations show that our system is much stronger than AES against channel noise perturbations, and thus can be better used for secure communications with large noise channel exists in open channels, e.g., mobile-phone secure communications.

nlin.CD↗

arXiv2003

A Spatiotemporal-chaos-based Encryption Having Overall Properties Considerably Better Than Advanced Encryption Standard

Spatiotemporal chaos of a two-dimensional one-way coupled map lattice is used for chaotic cryptography. The chaotic outputs of many space units are used for encryption simultaneously. This system shows satisfactory cryptographic properties of high security; fast encryption (decryption) speed; and robustness against noise disturbances in communication channel. The overall features of this spatiotemporal-chaos-based cryptosystem are better than chaotic cryptosystems known so far, and also than currently used conventional cryptosystems, such as the Advanced Encryption Standard (AES).

nlin.CD↗