Deep convolutional neural networks have achieved much success in remote sensing image change detection (CD) but still suffer from two main problems. First, existing multi-scale feature fusion methods often employ redundant extraction and strategies, which leads to high computational costs memory usage. Second, the regular attention mechanism CD is difficult model spatial-spectral features generate 3D weights at same time, ignoring cooperation between spatial spectral features. To address...

The popular Siamese convolutional neural networks (CNNs) for remote sensing (RS) image change detection (CD) often suffer from two problems. First, they either ignore the original information of bitemporal images or insufficiently utilize difference between images, which leads to low tightness changed objects. Second, CNNs always employ dual-branch encoders CD, increases computational cost. To address above issues, this article proposes a network based on enhancement and spatial–spectral...

With the development of earth observation technology, massive amounts remote sensing (RS) images are acquired. To find useful information from these images, cross-modal RS image-voice retrieval provides a new insight. This paper aims to study task so as search effective data. Existing methods for rely primarily on pairwise relationship narrow heterogeneous semantic gap between and voices. However, apart included in datasets, intra-modality non-paired inter-modality relationships should also...

Cross-modal biometric matching (CMBM) aims to determine the corresponding voice from a face, or identify face voice. Recently, many CMBM methods have been proposed by forcing distance between two modal features be narrowed. However, these ignore alignability features. Because feature is extracted under supervision of identity information single data, it can only reflect data. In order address this problem, disentangled representation learning method disentangle alignable latent factors and...

Haze always shrouds satellite images, obscuring valuable geographic information for military surveillance, natural calamity surveillance and mineral resource exploration. Satellite image dehazing (SID) provides the possibility better applications of images. Most existing methods are tailored images not very effective with non-homogeneous haze since semantic structure inconsistent attenuation fully considered. To tackle this problem, study proposes a hierarchical semantic-guided contextual...

Visual Attention Prediction (VAP) is a significant and imperative issue in the field of computer vision. Most existing VAP methods are based on deep learning. However, they do not fully take advantage low-level contrast features while generating visual attention map. In this paper, novel method proposed to generate map via bio-inspired representation The learning combines both high-level semantic simultaneously, which developed by fact that human eye sensitive patches with high objects...

Change detection is an important task of identifying changed information by comparing bitemporal images over the same geographical area. Currently, many existing methods based on U-Net and attention mechanism have greatly promoted development change techniques. However, they still suffer from two main challenges. First, faced with diversity ground objects flexibility scale changes, vanilla mechanisms cripple spatial in learning object details due to convolution kernels at different layers....

Abstract Aerial scene recognition (ASR) has attracted great attention due to its increasingly essential applications. Most of the ASR methods adopt multi‐scale architecture because both global and local features play roles in ASR. However, existing neglect effective interactions among different scales various spatial locations when fusing features, leading a limited ability deal with challenges large‐scale variation complex background aerial images. In addition, may suffer from poor...

Change Detection (CD) is the process of recognizing and quantitatively evaluating changes in surface objects at exact location but different times from remote sensing images that may be caused by extreme heat events. Attention-based CD methods have gained much traction because their ability to concentrate on change regions. However, most current attention-based only calculate attention matrix based features extracted a single image fail consider correlation times. The effectiveness learned...

Popular Transformer networks have been successfully applied to remote sensing (RS) image change detection (CD) identifications and achieved better results than most convolutional neural (CNNs), but they still suffer from two main problems. First, the computational complexity of grows quadratically with increase spatial resolution, which is unfavorable RS images. Second, these popular tend ignore importance fine-grained features, in poor edge integrity internal tightness for largely changed...

Popular Transformer networks have been successfully applied to remote sensing (RS) image change detection (CD) identifications and achieve better results than most convolutional neural (CNNs), but they still suffer from two main problems. First, the computational complexity of grows quadratically with increase spatial resolution, which is unfavorable very high-resolution (VHR) RS images. Second, these popular tend ignore importance fine-grained features, in poor edge integrity internal...

Haze or cloud always shrouds satellite images, obscuring valuable geographic information for military surveillance, natural calamity surveillance and mineral resource exploration. Satellite image dehazing (SID) provides the possibility better applications of images. Most existing methods are tailored images not very effective with non-homogeneous haze since semantic structure inconsistent attenuation fully considered. To tackle this problem, study proposes a hierarchical semantic-guided...

Remote-sensing image dehazing (RSID) is crucial for applications such as military surveillance and disaster assessment. However, current methods often rely on complex network architectures, compromising computational efficiency scalability. Furthermore, the scarcity of annotated remote-sensing-dehazing datasets hinders model development. To address these issues, a Dual-View Knowledge Transfer (DVKT) framework proposed to generate lightweight efficient student by distilling knowledge from...

Single-image super-resolution (SISR) reconstruction is important for image processing, and lots of algorithms based on deep convolutional neural network (CNN) have been proposed in recent years. Although these better accuracy recovery results than traditional methods without CNN, they ignore finer texture details when super-resolving at a large upscaling factor. To solve this problem, paper we propose an algorithm generative adversarial single-image restoration 4x factors. We decode restored...

When an image translation task contains intradomain translations, the untranslated source will be discriminated as real by discriminator. Thus if network’s nonlinearity is insufficient, generator can fool discriminator producing output that resembles image. We propose activation function termed “adaptive rectified linear unit (ReLU) with structure adaption (SA-AdaReLU)” to enhance control and of network in tasks. SA-AdaReLU composed two technologies: adaptive ReLU (AdaReLU) structural...