TABLE II: FVD and PSNR Comparison and a BD-rate Equivalent of Standard Test Sequences (“NN” refers to “nearest-neighbor”)įig. By using a discriminator and a mixed loss to guide the whole video compression network to optimize towards generating realistic decoded videos, our proposed DVC-P outperformed DVC in terms of a BD-rate equivalent and visual experience.
In this paper, we have proposed the DVC-P network aiming at restoring decoded videos in high perceptual quality at the decoder side. TABLE I: FVD and Bitrate Comparison of Standard Test Sequences at QP=37 ( λ = 256 and ) Iii-C2 Elimination of Checkerboard Artifacts In general, our proposed method can generate more realistic decoded videos and outperform DVC. 2 to compare the performance at 4 QPs, taking sequence RaceHorses(class D) as an example. In addition, We draw “FVD-Bit rate” curves in Fig. If we just focus on larger QPs (32 and 37), DVC-P still performs better. It is because on smaller QPs (22 and 27), where FVD is already low, the bit rate is higher. Notice that DVC-P performs worse on BQSquare.
We also compute a BD-rate equivalent (referred to “FVD BD-rate”) which indicates how much less bit rate the proposed method needs to achieve the same FVD as DVC for the same FVD, over 4 QP points: 22, 27, 32 and 37 (corresponding to λ = 2048, 1024, 512 and 256, ω = 1 / 1 2048 2048, 1 / 1 1024 1024, 1 / 1 512 512 and 1 / 1 256 256), as shown in Table II. Smaller FVD values correspond to better performance. λ = 256 corresponds to QP=37 in DVC), we compute FVD for all sequences at almost the same bit rate, as shown in Table I. When setting for proposed DVC-P and λ = 256 for DVC ( λ trades off between distortion and bit rate in DVC. We test perceptual quality of decoded videos by FVD.
#DVC CALCULATOR 2019 GENERATOR#
Ii-B Perceptual Optimizations Ii-B1 Proposed Generator and Discriminator In terms of estimating the bit rate, the entropy model in is used to calculate it. The motion compensation network achieves warp operation and prepares for residual calculation. It is fine tuned during the training process. A pretrained optical flow estimation network is used to estimate motion between the generated/reference frame and current raw frame. As for MV encoder network, its structure follows the same design as the residual encoder network. Since both are non-differentiable, during training quantization is replaced by additive uniform noise, and entropy coding is bypassed, approximating rate by the entropy of the latent representation. After quantization, the signal is losslessly processed by entropy coding to form the bit stream. There is a rectifier unit ( ReLu) after every convolution except the last one. Each layer downsamples its input with stride=2. Specifically, residual encoder network, which encodes residuals between the raw video frame and reconstructed video frame to bit streams, consists of four convolution layers. The structures of residual encoder network, motion vector (MV) encoder network, optical flow network, motion compensation network and bit rate estimation follow those in DVC network
Perceptual BD-rate equivalent, on average. Generate videos with higher perceptual quality achieving 12.27 Experimental resultsĭemonstrate that, compared with the baseline DVC, our proposed method can Perceptual quality of decoded sequences is improved. Interpolation is used to eliminate checkerboard artifacts which can appear in Off among distortion, perception and rate. Specifically, aĭiscriminator network and a mixed loss are employed to help our network trade Network, but improves it with perceptual optimizations. Our proposed DVC-P is based on Deep Video Compression (DVC) Perceptual optimizations (DVC-P), which aims at increasing perceptual quality In this paper, we introduce deep video compression with Video compression methods, which aim at optimizing objective or perceptual Recent years have witnessed the significant development of learning-based
0 kommentar(er)
