Multi-modal image registration aims to spatially align two images from different modalities make their feature points match with each other. Captured by sensors, the often contain many distinct features, which makes it challenging find accurate correspondences. With success of deep learning, networks have been proposed multi-modal images, however, they are mostly lack interpretability. In this paper, we first model problem as a disentangled convolutional sparse coding (DCSC) model. model,...

Nowadays, people are getting used to taking photos record their daily life, however, the actually not consistent with real natural scenes. The two main differences that tend have low dynamic range (LDR) and resolution (LR), due inherent imaging limitations of cameras. multi-exposure image fusion (MEF) super-resolution (SR) widely-used techniques address these issues. However, they usually treated as independent researches. In this paper, we propose a deep Coupled Feedback Network (CF-Net)...

Quality of experience (QoE) serves as a direct evaluation users' experiences in mobile video transmission and thus essential for network management, such optimization. In this paper, we propose deep learning-based QoE prediction approach with large-scale dataset transmission. Specifically, develop phone application collecting user data when viewing videos transmitted over the internet practical environment. Then, construct by 80000 piece four kinds subjective scores 89 parameters. Each...

Brain neural activity decoding is an important branch of neuroscience research and a key technology for the brain-computer interface (BCI). Researchers initially developed simple linear models machine learning algorithms to classify recognize brain activities. With great success deep on image recognition generation, networks (DNN) have been engaged in reconstructing visual stimuli from human via functional magnetic resonance imaging (fMRI). In this paper, we reviewed based algorithms....

Deep learning (DL) based semantic communication methods have been explored to transmit images efficiently in recent years. In this paper, we propose a generative model further improve the efficiency of image transmission and protect private information. particular, transmitter extracts interpretable latent representation from original by exploiting GAN inversion method. We also employ privacy filter knowledge base erase information replace it with natural features base. The simulation...

Recent studies have greatly promoted the development of semantic segmentation. Most state-of-the-art methods adopt fully convolutional networks (FCNs) to accomplish this task, in which connected layer is replaced with convolution for dense prediction. However, standard has limited ability maintaining continuity between predicted labels as well forcing local smooth. In paper, we propose unit (DCU), more suitable pixel-wise classification. The DCU adopts prediction instead center-prediction...

When watching omnidirectional images (ODIs), subjects can access different viewports by moving their heads. Therefore, it is necessary to predict subjects' head fixations on ODIs. Inspired generative adversarial imitation learning (GAIL), this paper proposes a novel approach saliency of ODIs, named SalGAIL. First, we establish dataset for attention ODIs (AOI). In contrast traditional datasets, our AOI large-scale, which contains the 30 viewing 600 Next, mine and discover three findings: (1)...

The existing image compression methods usually choose or optimize low-level representation manually. Actually, these struggle for the texture restoration at low bit rates. Recently, deep neural network (DNN)-based have achieved impressive results. To achieve better perceptual quality, generative models are widely used, especially adversarial networks (GAN). However, training GAN is intractable, high-resolution images, with challenges of unconvincing reconstructions and unstable training....

A synthetic aperture radar (SAR) imaging system usually produces pairs of bright area and dark when depicting the ground objects, such as a building or tree its shadow. Many buildings (trees) are aggregated together to form urban areas (forests). It means that often exist in scenes. Conventional unsupervised segmentation approaches segment scenes (e.g., forests) into different regions simply according gray values image. However, more convincing way is regard them consistent regions. In this...

With the dramatic growth of multimedia volume, semantic-oriented image representation and compression methods have proved to be important approaches improve efficiency in 6G scenarios. Semantic segmentation maps become carriers for semantic compressive coding due explicit description spatial categorical semantics core objects. This letter proposes an framework based on maps, which efficient flexible adjustment bit-rates are realized by controllable maps. Specifically, region-based map is...

Abstract This study examined the spatiotemporal characteristics of extreme maximum temperature events (EMTEs) in Central China using observational data from national meteorological stations and coupled model intercomparison project phase 6 models by focusing on variances between global warming thresholds 1.5, 2.0, 4°C. The threshold was determined based 99th percentile daily temperature, an improved intensity–area–duration method employed to determine EMTE characteristics. Results indicated...

Automated maritime object detection is a significant research challenge in intelligent marine surveillance systems for the Internet of Things (IoT) and smart ocean applications. In particular, ship recognized as one core issues these IoT-driven systems. Traditional methods based on machine learning have made some achievements tasks specific objects. However, objects are relatively small, they usually not accurately detected. this article, we propose an electroencephalography (EEG)-based...

The 360-degree video/image, also called an omnidirectional video/image or panoramic is very important in some emerging areas such as virtual reality (VR). Therefore, corresponding image generation algorithms are urgently needed. However, existing models mainly focus on 2-D images and do not consider the spherical structures of images. In this article, we propose a method based convolution generative adversarial networks, networks (SGANs). We adopt sketch map input, which concise geometric...

Although the bandwidth of high-resolution panchromatic (HR PAN) image is wide, it narrow in each band low-resolution multispectral (LR MS) image. Hence, spatial resolution HR PAN much higher than that LR MS However, only has a single band. The purpose Pan-sharpening algorithm to make Pan-sharpened with both high and good spectral information. In this paper, novel learning interpolation method for proposed by expanding sketch information contains edges lines features image, segment its own...

The latest High Efficiency Video Coding (HEVC) standard significantly improves coding efficiency over its previous video standards. expense of such improvement is enormous computational complexity, from both encoding and decoding sides. Since capability power capacity are diverse across portable devices, it necessary to reduce complexity a target with tolerable quality loss, so called control. This paper proposes saliency-guided control (SGCC) approach for HEVC decoding, which reduces the...

Compressing large images with a generative model goes beyond typical image encoding standards under notably low bitrate. In this paper, we step toward practical compression systems based on recent advances. Specifically, show that the channel redundancy of latent representation produced by an autoencoder network can be effectively compressed via mask compression. The performs quantization variance instead original values. Instead training multiple models, changing leads to simple and...