Benefiting from multi-view video plus depth and depth-image-based-rendering technologies, only limited views of a real 3-D scene need to be captured, compressed, transmitted. However, the quality assessment synthesized is very challenging, since some new types distortions, which are inherently different texture coding errors, inevitably produced by view synthesis map compression, corresponding original (reference views) usually not available. Thus full-reference metrics cannot used for...

Uncertainty is inherent in machine learning methods, especially those for camouflaged object detection aiming to finely segment the objects concealed background. The strong enquote center bias of training dataset leads models poor generalization ability as learn find around image center, which we define model bias. Further, due similar appearance and its surroundings, it difficult label accurate scope object, along boundaries, term data To effectively two types biases, resort uncertainty...

Understanding visual attention of observers on 360° images gains interest along with the booming trend Virtual Reality applications. Extending existing saliency prediction methods from traditional 2D to is not a direct approach due lack sufficient large image database. In this paper, we propose extend SalGAN, model based generative adversarial network, SalGAN360 by fine tuning SalGAN our new loss function predict both global and local maps. Our experiments show that outperforms tested...

With the growing data consumption of emerging video applications and users’ requirement for higher resolutions, up to 8K, a huge effort has been made in compression technologies. Recently, versatile coding (VVC) standardized by moving picture expert group (MPEG), providing significant improvement performance over its predecessor high efficiency (HEVC). In this paper, we provide comparative subjective quality evaluation between VVC HEVC standards 8K resolution videos. addition,...

The popularity of 3D applications, such as Free View-point TV (FTV) and Multi-view Video plus Depth (MVD), induces a heavy requirement synthesized views. However, the quality assessment views is very challenging because corresponding original (reference views) are usually not available at both encoder decoder sides. In this paper, we propose new no-reference model to evaluate views, called NIQSV (No-reference Image Quality Synthesized Views). This metric based on hypothesis that good image...

Nonlinear dynamic cryptosystems or chaos-based have been attracting a large amount of research since 1990. The critical aspect cryptography is to face the growth communication and achieve design fast secure cryptosystems. In this paper, we introduce three versions cryptosystem based on similar structure Zhang Fridrich Each version composed two layers: confusion layer diffusion layer. achieved by using modified 2-D cat map overcome fixed-point problem some other weaknesses, also increase key...

Depth-image-based rendering (DIBR) is a fundamental technology in several 3-D-related applications, such as free viewpoint video, virtual reality, and augmented reality. However, new challenges have also been brought assessing the quality of DIBR-synthesized views since this process induces some types distortions, which are inherently different from distortion caused by video coding. In paper, we present image database with associated subjective scores. We test performances state-of-the-art...

Given their outstanding performance, the Deep Neural Networks (DNNs) models have been deployed in many real-world applications. However, recent studies demonstrated that they are vulnerable to small carefully crafted perturbations, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> , adversarial examples, which considerably decrease performance and can lead devastating consequences, especially for safety-critical applications, such as...

Two High Efficiency Video Coding (HEVC) extensions, namely, the scalable HEVC (SHVC) extension and multiview (MV-HEVC) extension, have been finalized in July 2014 by Moving Picture Experts Group Group. These two extensions enable additional features not covered first version of standard such as spatial, fidelity, bitdepth, color gamut scalability, well stereoscopic representations. In this paper, we propose a software parallel decoder architecture for its multilayer including SHVC MV-HEVC...

Video compression and content quality have become one of the most research topic in recent years. Predominantly, trends obviously signpost that video usage over Internet is on upsurge. Simultaneously, users' requirement for enlarged resolution higher rising. Consequently, a huge effort has been made coding technologies monitoring. In this paper, we present subjective-based comparison as well an objective measurement between newest Versatile Coding (VVC) well-known High Efficiency (HEVC)...

Ambisonics, which constructs a sound distribution over the full viewing sphere, improves immersive experience in omnidirectional video (ODV) by enabling observers to perceive directions. Thus, human attention could be guided audio and visual stimuli simultaneously. Numerous datasets have been proposed investigate collecting eye fixations of navigating ODV with head-mounted displays (HMD). However, there is no such dataset analyzing impact information. In this paper, we establish new...

The rise of deep learning (DL) has increased computing complexity and energy use, prompting the adoption application specific integrated circuits (ASICs) for energy-efficient edge mobile deployment. However, recent studies have demonstrated vulnerability these accelerators to attacks. Despite development various inference time attacks in prior research, backdoor remain unexplored. In this paper, we design an innovative attack against neural networks (DNNs) operating on sparsity-based...

Overfitted neural video codecs offer a decoding complexity orders of magnitude smaller than their autoencoder counterparts. Yet, this low comes at the cost limited compression efficiency, in part due to difficulty capturing accurate motion information. This paper proposes guide information learning with an optical flow estimator. A joint rate-distortion optimization is also introduced improve rate distribution across different frames. These contributions maintain 1300 multiplications per...

Computer vision applications such as refocusing, segmentation and classification become one of the most advanced imaging services. Light Field (LF) systems provide a rich semantic information scene. Using dense set cameras microlens arrays (Plenoptic camera), direction each ray coming from scene toward LF capture system can be extracted represented by spatial angular coordinates. However, induces many drawbacks including large amount data produced complexity increase for representation. In...

The future video coding standard named Versatile Video Coding (VVC) is expected by the end of 2020. VVC will enable better efficiency than current High Efficiency (HEVC) standard. This gain brought several tools. Multiple Transform Selection (MTS) one key tools that have been introduced in VVC. MTS concept relies on three transform types including Discrete Cosine (DCT)-II, Sine (DST)-VII and DCT-VIII. Unlike DCT-II has fast computing algorithms, DST-VII DCT-VIII rely more complex matrix...

Light field technology represents a viable path for providing high-quality VR content. However, such an imaging system generates high amount of data leading to urgent need LF image compression solution. In this paper, we propose efficient coding scheme based on view synthesis. Instead transmitting all the views, only some them are coded and transmitted, while remaining views dropped. The transmitted using Versatile Video Coding (VVC) used as reference synthesize missing at decoder side....

This paper introduces a new method for inter-frame coding based on two complementary autoencoders: MOFNet and CodecNet. aims at computing conveying the Optical Flow pixel-wise Mode selection. The optical flow is used to perform prediction of frame code. mode selection enables competition between direct copy or transmission through proposed scheme assessed under Challenge Learned Image Compression 2020 (CLIC20) P-frame conditions, where it shown par with state-of-the-art video codec ITU/MPEG...

This article provides an overview of the main objectives new RVC standard, with emphasis on features that enable efficient implementation platforms multiple cores. A brief introduction to methodologies efficiently map codec specifications multicore is accompanied example possible breakthroughs are expected occur in design and deployment multimedia services platforms.

Deep neural networks (DNNs) have recently achieved state-of-the-art performance and provide significant progress in many machine learning tasks, such as image classification, speech processing, natural language etc. However, recent studies shown that DNNs are vulnerable to adversarial attacks. For instance, the classification domain, adding small imperceptible perturbations input is sufficient fool DNN cause misclassification. The perturbed image, called example, should be visually close...

In this paper, we propose, implement, and analyze the structures of two keyed hash functions using Chaotic Neural Network (CNN). These are based on Sponge construction, they produce variants value lengths, i.e., 256 512 bits. The first structure is composed two-layered CNN, while second one formed by one-layered CNN a combination nonlinear functions. Indeed, proposed employ strong systems, precisely chaotic system neural network system. addition, study new methodology combining networks...