Generative adversarial network (GANs) is one of the most important research avenues in field artificial intelligence, and its outstanding data generation capacity has received wide attention. In this paper, we present recent progress on GANs. First, basic theory GANs differences among different generative models years were analyzed summarized. Then, derived are classified introduced by one. Third, training tricks evaluation metrics given. Fourth, applications introduced. Finally, problem,...

The high definition (HD) and ultra HD videos can be widely applied in broadcasting applications. However, with the increased resolution of video, volume raw visual information data increases significantly, which becomes a challenge for storage, processing, transmitting data. state-of-the-art video compression standard-H.265/High Efficiency Video Coding (HEVC) compresses efficiently, while rate comes at cost heavy computation load. Hence, reducing encoding complexity vital H.265/HEVC encoder...

Guided by the free-energy principle, generative adversarial networks (GAN)-based no-reference image quality assessment (NR-IQA) methods have improved prediction accuracy. However, GAN cannot well handle restoration task for principle-guided NR-IQA methods, especially severely destroyed images, which results in that reconstruction relationship between distorted and its restored be accurately built. To address this problem, a visual compensation network (VCRNet)-based method is proposed, uses...

The high efficiency video coding (HEVC) is the latest standard, which adopts quadtree structure based tree unit (CTU) to improve efficiency. In HEVC encoding process, CTU recursively split into 8×8 size units (CUs) from 64×64 CU. Along with increased number of sizes CUs, modes has been greatly increased, results in computational complexity encoder. this paper, we propose an early MERGE mode decision algorithm reduce Firstly, on all-zero block (AZB) and motion estimation (ME) information...

The raw video data can be compressed much by the latest coding standard, high efficiency (HEVC). However, block-based hybrid used in HEVC will incur lots of artifacts videos, quality severely influenced. To settle this problem, in-loop filtering is to eliminate artifacts. Inspired success deep learning, we propose an efficient algorithm based on enhanced convolutional neural networks (EDCNN) for significantly improving performance HEVC. Firstly, problems traditional models, including...

The Versatile Video Coding (VVC) achieves superior coding efficiency as compared with the High Efficiency (HEVC), while its excellent performance is at cost of several high computational complexity tools, such Quad-Tree plus Multi-type Tree (QTMT)-based Units (CUs) and multiple inter prediction modes. To reduce VVC, a CNN-based fast method proposed in this paper. First, multi-information fusion CNN (MF-CNN) model to early terminate QTMT-based CU partition process by jointly using...

Deep neural networks have achieved great performance on blind Image Quality Assessment (IQA), but it is still challenging for using one network to accurately predict the quality of images with different distortions. In this paper, a Distortion-Aware Convolutional Neural Network (DACNN) proposed IQA, which works effectively not only synthetically distorted also authentically images. The DACNN consists distortion aware module, fusion and prediction module. Siamese network-based pretraining...

Screen content coding (SCC) has evolved into the extension of High Efficiency Video Coding (HEVC). Low-latency, real-time transport between devices in form screen video is becoming popular many applications. However, complexity encoder still very high for intra prediction HEVC-based SCC. This paper proposes a fast method based on property analysis First, units (CUs) are classified natural CUs (NCCUs) and (SCCUs), statistic characteristics content. For NCCUs, newly adopted modes, including...

High Efficiency Video Coding (HEVC) improves the compression efficiency at cost of high computational complexity by using quad-tree coding unit (CU) structure and variable prediction (PU) modes. To minimize HEVC encoding while maintaining its efficiency, a binary multi-class support vector machine (SVM)-based fast algorithm is presented in this paper. First, processes recursive CU decision PU selection are modeled as hierarchical classification structures. Second, according to two...

Recently, the Generative Adversarial Networks (GANs) are fast becoming a key promising research direction in computational intelligence. To improve modeling ability of GANs, loss functions used to measure differences between samples generated by model and real samples, make learn towards goal. In this paper, we perform survey for analyze pros cons these functions. Firstly, basic theory its training mechanism introduced. Then, GANs summarized, including not only objective but also...

In three-dimensional video system, the texture and depth videos are jointly encoded, then Depth Image Based Rendering (DIBR) is utilized to realize view synthesis. However, compression distortion of videos, as well disocclusion problem in DIBR degrade visual quality synthesized view. To address this problem, a Two-stream Attention Network (TSAN)-based enhancement method proposed for 3D-High Efficiency Video Coding (3D-HEVC) article. First, shortcomings synthesis technique traditional...

No-reference image quality assessment (NR-IQA) aims to evaluate without using the original reference images. Since early NR-IQA methods based on distortion types were only applicable specific scenarios, and lack of practicality, it is challenging designing a universal method. In this article, multibranch convolutional neural network (MB-CNN) method proposed, which includes spatial-domain feature extractor, gradient-domain weight mechanism. The extractor extract features from spatial domain....

High Efficiency Video Coding (HEVC) INTRA coding improves compression efficiency by adopting advanced technologies, such as multi-level quad-tree block partitioning and up to 35-mode prediction. However, it significantly increases the complexity, memory access, power consumption, which goes against its widely applications, especially for ultra-high definition and/or mobile video applications. To tackle this problem, we propose effective data driven unit (CU) size decision approaches HEVC...

High-Efficiency Video Coding (HEVC) efficiently addresses the storage and transmit problems of high-definition videos, especially for 4K videos. The variable-size Prediction Units (PUs)--based Motion Estimation (ME) contributes a significant compression rate to HEVC encoder also generates huge computation load. Meanwhile, high-level encoding complexity prevents widespread adoption in multimedia systems. In this article, an adaptive fractional-pixel ME skipped scheme is proposed...

Visual quality of images captured by mobile devices is often inferior to that a Digital Single Lens Reflex (DSLR) camera. This paper presents novel generative adversarial network-based image enhancement method, referred as MIEGAN. It consists multi-module cascade network and adaptive multi-scale discriminative network. The built upon two-stream encoder, feature transformer, decoder. In the luminance-regularizing stream proposed help focus on low-light areas. transformation module, two...

Stereopsis is the ability of human beings to get 3D perception on real scenarios. The conventional stereopsis measurement based subjective judgment for stereograms, leading be easily affected by personal consciousness. To alleviate issue, in this paper, EEG signals evoked dynamic random dot stereograms (DRDS) are collected stereogram recognition, which can help ophthalmologists diagnose strabismus patients even without real-time communication. classify signals, a novel multi-scale temporal...

In the three-dimensional video system, depth image-based rendering is a key technique for generating synthesized views, which provides audiences with perception and interactivity. However, inaccuracy of information leads to geometrical position errors, compression distortion texture videos degrades quality views. Although existing enhancement methods can eliminate distortions in their huge computational complexity hinders applications real-time multimedia systems. To this end, residual...

INTRA video coding is essential for high quality mobile communication and industrial applications since it enhances quality, prevents error propagation, facilitates random access. The latest high-efficiency (HEVC) standard has adopted flexible quad-tree-based block structure complex angular prediction to improve the efficiency. However, these technologies increase complexity significantly, which consumes large hardware resources, computing time power cost, an obstacle real-time applications....

Abstract Healthcare data analysis is currently a challenging and crucial research issue for the development of robust disease diagnosis prediction system. Many specific few common methods have been discussed in literature healthcare classification. The present study implements 32 classification six categories (Bayes, function‐based, lazy, meta, rule‐based, tree‐based) with objective searching best mining. performance each method has evaluated based on time, accuracy, precision, recall,...

Among the existing video-related applications, a large proportion have requirements for scalability of video coding complexity, such as live chatting and on power-limited mobile devices. Hence, complexity control algorithms, which aim to make an effective flexible tradeoff between rate-distortion (RD) performance, great practical value. In this paper, novel scheme high efficiency (HEVC) is proposed by dynamically adjusting depth range each tree unit (CTU). To accurately, statistical model...

Generation of a 3D model an object from multiple views has wide range applications. Different parts would be accurately captured by particular view or subset in the case views. In this paper, novel coarse-to-fine network (C2FNet) is proposed for point cloud generation C2FNet generates subsets points that are best individual with support other way, and then fuses these to whole cloud. It consists coarse module where clouds constructed exploring cross-view spatial relations, fine features...

Image inpainting is a significant task in the applications of computer vision, that aims to fill damaged regions with visually realistic contents. With development deep learning, generative adversarial network (GAN)-based image approaches have achieved remarkable progress. However, these methods only utilize one-sided structure information assist inpainting, which can not achieve satisfactory results, especially when synthesizing large-area missing complex images. In order tackle this...

The traditional image steganography approaches usually need a cover so that the secret information can be imperceptibly embedded into it for communication. However, by utilizing embedding traces left in image, existence of could successfully determined steganalysis methods. To resist existing methods, coverless approach is proposed using histograms oriented gradients (HOGs)-based hashing algorithm. More specially, instead designating embedding, original images whose hash sequences equal to...