High Efficiency Video Coding (HEVC) provides superior coding efficiency than previous video standards at the cost of increasing encoding complexity. The complexity increase motion estimation (ME) procedure is rather significant, especially when considering complicated partitioning structure HEVC. To fully exploit brought by HEVC requires a huge amount computations. In this paper, we analyze ME in and propose parallel framework to decouple for different partitions on many-core processors....

High Efficiency Video Coding (HEVC) uses a very flexible tree structure to organize coding units, which leads superior efficiency compared with previous video standards. However, such unit also places great challenge for encoders. In order fully exploit the brought by this structure, huge amount of computational complexity is needed an encoder decide optimal each image block. One way achieve use parallel computing enabled many-core processors. paper, we analyze processors make decision....

Existing methods for single image super-resolution (SR) are typically evaluated with synthetic degradation models such as bicubic or Gaussian downsampling. In this paper, we investigate SR from the perspective of camera lenses, named CameraSR, which aims to alleviate intrinsic tradeoff between resolution (R) and field-of-view (V) in realistic imaging systems. Specifically, view R-V a latent model process learn reverse it low- high-resolution pairs. To obtain paired images, propose two novel...

We propose a pseudo-sequence-based scheme for light field image compression. In our scheme, the raw captured by camera is decomposed into multiple views according to lenslet array of that camera. These constitute pseudo sequence like video, and redundancy between exploited video encoder. The specific coding order views, prediction structure, rate allocation have been investigated encoding sequence. Experimental results show superior performance which achieves as high 6.6 dB gain compared...

We present a novel 2-D wavelet transform scheme of adaptive directional lifting (ADL) in image coding. Instead alternately applying horizontal and vertical lifting, as practice, ADL performs lifting-based prediction local windows the direction high pixel correlation. Hence, it adapts far better to orientation features windows. The is achieved by existing 1-D wavelets seamlessly integrated into global transform. predicting updating signals can be derived even at fractional precision level...

Inspired by the recent advances of image super-resolution using convolutional neural network (CNN), we propose a CNN-based block up-sampling scheme for intra frame coding. A can be down-sampled before being compressed normal coding, and then up-sampled to its original resolution. Different from previous studies on down/up-sampling-based methods in our have been designed training CNN instead hand-crafted. We explore new structure up-sampling, which features deconvolution feature maps,...

We study the dual problem of image super-resolution (SR), which we term compact-resolution (CR). Opposite to SR that hallucinates a visually plausible high-resolution given low-resolution input, CR provides version image, such is both pleasing and as informative possible compared image. propose convolutional neural network (CNN) for CR, namely, CNN-CR, inspired by great success CNN SR. Specifically, translate requirements into operable optimization targets training CNN-CR: visual quality...

This paper proposes a robust single-image super-resolution method for enlarging low quality web image/video degraded by downsampling and compression. To simultaneously improve the resolution perceptual of such image/video, we bring forward practical solution which combines adaptive regularization learning-based super-resolution. The contribution this work is twofold. First, propose to analyze image energy change characteristics during iterative process, i.e., ratio between primitive (e.g.,...

One key challenge to learning-based video compression is that motion predictive coding, a very effective tool for compression, can hardly be trained into neural network. In this paper, we propose the concept of Pixel-MotionCNN (PMCNN) which includes extension and hybrid prediction networks. PMCNN model spatiotemporal coherence effectively perform coding inside learning On basis PMCNN, further explore framework with additional components iterative analysis/synthesis binarization. The...

To effectively reduce JPEG compression artifacts, we propose a deep convolutional sparse coding (DCSC) network architecture. We design our DCSC in the framework of classic learned iterative shrinkage-threshold algorithm. focus on recognizing and separating artifacts only, sparsely code feature maps instead raw image. The final de-blocked image is directly reconstructed from coded features. use dilated convolution to extract multi-scale features, which allows single model simultaneously...

The 3D extension of High Efficiency Video Coding significantly improves the coding efficiency video at expense computational complexity. This paper presents a novel fast mode decision algorithm for depth map based on grayscale similarity and inter-view correlation. First, is adopted to judge whether reference frame could assist current frame. When difference in average between co-located unit (CU) CU smaller than threshold, level will be restricted by that coded CU. Second, correlation are...

In this paper, we study a simplified affine motion model-based coding framework to overcome the limitation of translational model and maintain low-computational complexity. The proposed mainly has three key contributions. First, propose reduce number parameters from 6 4. four-parameter can not only handle most complex motions in natural videos, but also save bits for two parameters. Second, efficiently encode parameters, prediction modes, i.e., an advanced vector scheme combined with...

We propose a Deep Boosting Framework (DBF) for real-world image denoising by integrating the deep learning technique into boosting algorithm. The DBF replaces conventional handcrafted units elaborate convolutional neural networks, which brings notable advantages in terms of both performance and speed. design lightweight Dense Dilated Fusion Network (DDFN) as an embodiment unit, addresses vanishing gradients during training due to cascading networks while promoting efficiency limited...

Diabetic retinopathy (DR) is the leading cause of permanent blindness in working-age population. And automatic DR diagnosis can assist ophthalmologists to design tailored treatments for patients, including grading and lesion discovery. However, most existing methods treat discovery as two independent tasks, which require annotations a learning guidance limits actual deployment. To alleviate this problem, we propose novel lesion-aware transformer (LAT) jointly unified deep model via an...

Wavelet transform is a powerful tool for multiresolution time-frequency analysis. It has been widely adopted in many image processing tasks, such as denoising, enhancement, fusion, and especially compression. Wavelets lead to the successful coding standard JPEG-2000. Traditionally, wavelets were designed from signal theory with certain assumption on signal, but natural images are not ideal assumed by theory. How design content-adaptive remains difficulty. Inspired recent progress of...

Single image denoising suffers from limited data collection within a noisy image. In this paper, we propose novel scheme, which explores both internal and external correlations with the help of web images. For each patch, build cubes by finding similar patches images, respectively. We then reducing noise two-stage strategy using different filtering approaches. first stage, since patch may lead to inaccurate selection, graph based optimization method improve matching accuracy in denoising....

We observe two trends, growing wireless capability at the physical layer powered by MIMO-OFDM, and video traffic as dominant application traffic. Both source MIMO-OFDM channel components exhibit non-uniform energy distribution. This motivates us to leverage data redundancy achieve high recovery performance. propose ParCast that first separates into independent components, matches more important with higher-gain allocates power weights joint consideration channel, uses analog modulation for...

We propose a novel dual-camera design to acquire 4D high-speed hyperspectral (HSHS) videos with high spatial and spectral resolution. Our work has two key technical contributions. First, we build system that simultaneously captures panchromatic video at frame rate low rate, which jointly provide reliable projections for the underlying HSHS video. Second, exploit learn an over-complete 3D dictionary represent each band-wise sparsely, robust computational reconstruction is then employed...

Fractional-pixel motion compensation (MC) improves the efficiency of inter prediction and has been utilized extensively in video coding standards. The traditional methods fractional-pixel MC usually follow approach interpolation, i.e., they adopt different kinds filters, either fixed or adaptive, to interpolate values from integer-pixel a reference picture. Different interpolation approach, this paper, we formulate as an inter-picture regression problem, which is predict pixel current...

Phase shifting profilometry (PSP) and Fourier transform (FTP) are two well-known fringe analysis methods for 3D sensing. PSP offers high accuracy but requires multiple images; FTP uses a single image is limited in its accuracy. In this paper, we propose novel Fourier-assisted phase (FAPS) method accurate dynamic Our key observation that the motion vulnerability of multi-shot can be overcome through single-shot FTP, while preserved. Moreover, to solve ambiguity complex scenes without...

In both H.264 and HEVC, context-adaptive binary arithmetic coding (CABAC) is adopted as the entropy method. CABAC relies on manually designed binarization processes well handcrafted context models, which may restrict compression efficiency. this paper, we propose an strategy by training neural networks, make preliminary studies of intra prediction modes in HEVC. Instead binarization, to directly estimate probability distribution 35 with adoption a multi-level codec. utilize convolutional...