“Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告

The PASCAL Visual Object Classes Challenge

The compression capability of several generations of video coding standards is compared by means of peak signal-to-noise ratio (PSNR) and subjective testing results. A unified approach is applied to the analysis of designs, including H.262/MPEG-2 Video, H.263, MPEG-4 Visual, H.264/MPEG-4 Advanced Video Coding (AVC), and High Efficiency Video Coding (HEVC). The results of subjective tests for WVGA and HD sequences indicate that HEVC encoders can achieve equivalent subjective reproduction quality as encoders that conform to H.264/MPEG-4 AVC when using approximately 50% less bit rate on average. The HEVC design is shown to be especially effective for low bit rates, high-resolution video content, and low-delay communication applications. The measured subjective improvement somewhat exceeds the improvement measured by the PSNR metric.

http://amalia.img.lx.it.pt/~fp/cav/Additional_material/HEVC-Performance.pdf

Comparison of the Coding Efficiency of Video Coding Standards—Including High Efficiency Video Coding (HEVC)

We present DeblurGAN, an end-to-end learned method for motion deblurring. The learning is based on a conditional GAN and the content loss. DeblurGAN achieves state-of-the art performance both in the structural similarity measure and visual appearance. The quality of the deblurring model is also evaluated in a novel way on a real-world problem - object detection on (de-)blurred images. The method is 5 times faster than the closest competitor - Deep-Deblur [25]. We also introduce a novel method for generating synthetic motion blurred images from sharp ones, allowing realistic dataset augmentation. The model, code and the dataset are available at https://github.com/KupynOrest/DeblurGAN

/pdf/deblurgan-blind-motion-deblurring-using-conditional-150wx51y9w.pdf

DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks

We present a comprehensive study and evaluation of existing single-image dehazing algorithms, using a new large-scale benchmark consisting of both synthetic and real-world hazy images, called REalistic Single-Image DEhazing (RESIDE). RESIDE highlights diverse data sources and image contents, and is divided into five subsets, each serving different training or evaluation purposes. We further provide a rich variety of criteria for dehazing algorithm evaluation, ranging from full-reference metrics to no-reference metrics and to subjective evaluation, and the novel task-driven evaluation. Experiments on RESIDE shed light on the comparisons and limitations of the state-of-the-art dehazing algorithms, and suggest promising future directions.

Benchmarking Single-Image Dehazing and Beyond

A method ofvisual media processing includes performing a conversion between a current block of visual media data and a corresponding coded representation of the current video block, wherein the conversion of the current block includes computing a spatial gradient based on sample differences of spatially neighboring samples and/or a temporal gradient based on temporally neighboring samples associated with the current block, and wherein a technique used for computing the spatial gradient is same as a technique used for computing the temporal gradient.

Unified calculation method for inter prediction with refinement

A method for visual mediaprocessing, comprising: computing, during a conversion between a current video block of visual media data and a bitstream representation of the current video block, a cross-component linear model (CCLM) and/or a chroma residual scaling (CRS) factor for the current video block based, at least in part, on neighboring samples of a corresponding luma block which covers a top-left sample of a collocated luma blockassociated with the current video block, wherein one or more characteristics of the current video block are used for identifying the corresponding luma block.

Methods for cross component dependency reduction

The application relates to improving method for transform or quantization bypass mode. A method of video processing is disclosed. The method of video processing includes: determining, during a conversion between a current block and a bitstream representation of the current block, that the current block is coded in a transquant bypass coding mode; and performing the conversion between the current block and a bitstream representation of the current block without filtering based on an adaptive loop filter in response to the current block being coded in the transquant bypass coding mode, wherein the transquant bypass coding mode is a coding mode in which a block is coded without one or more of a transform, a quantization and an in-loop filtering.

Improving method for transform or quantization bypass mode

A method of video processing is provided to comprise: maintaining, prior to a conversion between a current video block of a video region and a coded representation of the video, at least one history-based motion vector prediction (HMVP) table, wherein the HMVP table includes one or more entries corresponding to motion information of one or more previously processed blocks; and performing the conversion using the at least one HMVP table; and wherein the motion information of each entry is configured to include interpolation filter information for the one or more previously processed blocks, wherein the interpolation filter information indicates interpolation filters used for interpolating prediction blocks of the one or more previously processed blocks.

Using interpolation filters for history based motion vector prediction

A method of visual media processing, comprising: determining, for a current video block, a motion vector for use in a sub-block based motion vector prediction (sbTMVP) process to locate a corresponding block in a collocated picture for a conversion between the current video block and a bitstream representation of the current video block, wherein the motion vector used in the sbTMVP process is computed in accordance with a scaling operation; and performing, based on using the motion vector, the conversion between the current video block and the bitstream representation of the visual media data.

Xu Jizheng

Papers

Unified calculation method for inter prediction with refinement

Methods for cross component dependency reduction

Improving method for transform or quantization bypass mode

Using interpolation filters for history based motion vector prediction

Transform unit based combined inter intra prediction