Evgeny Belyaev

Dr. Evgeny Belyaev

ITMO University, Kronverksky prospekt 49, 197101 Saint-Petersburg, Russia
orcid

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
15 February 2022 in Engineering Applications of Artificial Intelligence
Verified
Journal Article
Closed access
CDChongyang Ding
Chongyang Ding
SWShan Wen
Shan Wen
WDWenwen Ding
Wenwen Ding
KLKai Liu
Kai Liu
Evgeny BelyaevEvgeny Belyaev
Evgeny Belyaev

10 July 2020 in IEEE Access
Verified
Journal Article
Open access
LYLin Yan
Lin Yan
KLKai Liu
Kai Liu
Evgeny BelyaevEvgeny Belyaev
Evgeny Belyaev
The limitation of LiDAR (Light Detection And Ranging) sensor causes the general sparsity of produced depth measurement. However, the sparse representation of the world is insufficient for applications such as 3D reconstruction. Thus, depth completion is an important computer vision task in which a synchronized RGB image is commonly available. In this paper, we propose a deep neural network to tackle this image guided depth completion problem. By revisiting the sparsity invariant convolution and revealing how it can be used in a novel approach, we propose three mask aware operations to process, downscale, and fuse sparse features. These operations explicitly consider the observation mask of its corresponding feature map. In addition, the structure of this network follows a novel scheme in which data from image and depth domain are processed by these proposed operations independently. Our proposed model achieves state-of-the-art performance on the KITTI depth completion benchmark. Furthermore, it presents a strong robustness for bearing input sparsity under different densities and patterns.

26 January 2023 in Sensors
Verified
Journal Article
Open access
PDF file
Evgeny BelyaevEvgeny Belyaev
Evgeny Belyaev
This paper is dedicated to video coding based on a compressive sensing (CS) framework. In CS, it is assumed that if a video sequence is sparse in some transform domain, then it could be reconstructed from a much lower number of samples (called measurements) than the Nyquist–Shannon theorem requires. Here, the performance of such a codec depends on how the measurements are acquired (or sensed) and compressed and how the video is reconstructed from the decoded measurements. Here, such a codec potentially could provide significantly faster encoding compared with traditional block-based intra-frame encoding via Motion JPEG (MJPEG), H.264/AVC or H.265/HEVC standards. However, existing video codecs based on CS are inferior to the traditional codecs in rate distortion performance, which makes them useless in practical scenarios. In this paper, we present a video codec based on CS called CS-JPEG. To the author’s knowledge, CS-JPEG is the first codec based on CS, combining fast encoding and high rate distortion results. Our performance evaluation shows that, compared with the optimized software implementations of MJPEG, H.264/AVC, and H.265/HEVC, the proposed CS-JPEG encoding is 2.2, 1.9, and 30.5 times faster, providing 2.33, 0.79, and 1.45 dB improvements in the peak signal-to-noise ratio, respectively. Therefore, it could be more attractive for video applications having critical limitations in computational resources or a battery lifetime of an upstreaming device.

10 April 2020 in IET Image Processing
Verified
Journal Article
Open access
PDF file
Evgeny BelyaevEvgeny Belyaev
Evgeny Belyaev
MCMarian Codreanu
Marian Codreanu
MJMarkku Juntti
Markku Juntti
KEKaren Egiazarian
Karen Egiazarian
The authors consider the problem of compressive sensed video recovery via iterative thresholding algorithm. Traditionally, it is assumed that some fixed sparsifying transform is applied at each iteration of the algorithm. In order to improve the recovery performance, at each iteration the thresholding could be applied for different transforms in order to obtain several estimates for each pixel. Then the resulting pixel value is computed based on obtained estimates using simple averaging. However, calculation of the estimates leads to significant increase in reconstruction complexity. Therefore, the authors propose a heuristic approach, where at each iteration only one transform is randomly selected from some set of transforms. First, they present simple examples, when block‐based 2D discrete cosine transform is used as the sparsifying transform, and show that the random selection of the block size at each iteration significantly outperforms the case when fixed block size is used. Second, building on these simple examples, they apply the proposed approach when video block‐matching and 3D filtering (VBM3D) is used for the thresholding and show that the random transform selection within VBM3D allows to improve the recovery performance as compared with the recovery based on VBM3D with fixed transform.

17 May 2017 in IEEE Signal Processing Letters
Verified
Journal Article
Closed access
Evgeny BelyaevEvgeny Belyaev
Evgeny Belyaev
SFSøren Forchhammer
Søren Forchhammer
KLKai Liu
Kai Liu
We propose a novel efficient multialphabet multiplication-free adaptive arithmetic coder. First, we generalize probability estimation via virtual sliding window for the multialphabet case and show that it does not require multiplications and provides a tradeoff between the probability adaptation speed and the precision of the probability estimation. Second, we show how the generalized virtual sliding window can be used to eliminate multiplications and divisions. Finally, we demonstrate that the proposed arithmetic coder provides better compression performance than existing implementations based on state-of-the-art multiplication-free binary arithmetic coders.

24 April 2017 in IEEE Communications Letters
Verified
Journal Article
Closed access
Evgeny BelyaevEvgeny Belyaev
Evgeny Belyaev
SFSøren Forchhammer
Søren Forchhammer
MCMarian Codreanu
Marian Codreanu
Error concealment for video coding based on a 3-D discrete wavelet transform (DWT) is considered. We assume that the video sequence has a sparse representation in a known basis different from the DWT, e.g., in a 2-D discrete cosine transform basis. Then, we formulate the concealment problem as l 1 -norm minimization and solve it utilizing an iterative thresholding algorithm. Comparing different thresholding operators, we show that video block-matching and 3-D filtering provide the best reconstruction by utilizing spatial similarity within a frame and temporal similarity between neighbor frames. Experimental results show that the proposed error concealment provides up to 6-dB improvement in the peak signal-to-noise ratio in comparison with existing recovery methods.