Showing papers on "Entropy encoding published in 2018"

Improved Lossy Image Compression with Priming and Spatially Adaptive Bit Rates for Recurrent Networks

Abstract: We propose a method for lossy image compression based on recurrent, convolutional neural networks that outperforms BPG (4:2:0), WebP, JPEG2000, and JPEG as measured by MS-SSIM. We introduce three improvements over previous research that lead to this state-of-the-art result using a single model. First, we modify the recurrent architecture to improve spatial diffusion, which allows the network to more effectively capture and propagate image information through the network's hidden state. Second, in addition to lossless entropy coding, we use a spatially adaptive bit allocation algorithm to more efficiently use the limited number of bits to encode visually complex image regions. Finally, we show that training with a pixel-wise loss weighted by SSIM increases reconstruction quality according to multiple metrics. We evaluate our method on the Kodak and Tecnick image sets and compare against standard codecs as well as recently published methods based on deep neural networks.

Practical Full Resolution Learned Lossless Image Compression

Abstract: We propose the first practical learned lossless image compression system, L3C, and show that it outperforms the popular engineered codecs, PNG, WebP and JPEG 2000. At the core of our method is a fully parallelizable hierarchical probabilistic model for adaptive entropy coding which is optimized end-to-end for the compression task. In contrast to recent autoregressive discrete probabilistic models such as PixelCNN, our method i) models the image distribution jointly with learned auxiliary representations instead of exclusively modeling the image distribution in RGB space, and ii) only requires three forward-passes to predict all pixel probabilities instead of one for each pixel. As a result, L3C obtains over two orders of magnitude speedups when sampling compared to the fastest PixelCNN variant (Multiscale-PixelCNN). Furthermore, we find that learning the auxiliary representation is crucial and outperforms predefined auxiliary representations such as an RGB pyramid significantly.

Universal Deep Neural Network Compression

Abstract: In this paper, we investigate lossy compression of deep neural networks (DNNs) by weight quantization and lossless source coding for memory-efficient deployment. Whereas the previous work addressed non-universal scalar quantization and entropy coding of DNN weights, we for the first time introduce universal DNN compression by universal vector quantization and universal source coding. In particular, we examine universal randomized lattice quantization of DNNs, which randomizes DNN weights by uniform random dithering before lattice quantization and can perform near-optimally on any source without relying on knowledge of its probability distribution. Moreover, we present a method of fine-tuning vector quantized DNNs to recover the performance loss after quantization. Our experimental results show that the proposed universal DNN compression scheme compresses the 32-layer ResNet (trained on CIFAR-10) and the AlexNet (trained on ImageNet) with compression ratios of $47.1$ and $42.5$, respectively.

End-To-End Optimized Speech Coding with Deep Neural Networks

Abstract: Modern compression algorithms are often the result of laborious domain-specific research; industry standards such as MP3, JPEG, and AMR-WB took years to develop and were largely hand-designed. We present a deep neural network model which optimizes all the steps of a wideband speech coding pipeline (compression, quantization, entropy coding, and decompression) end-to-end directly from raw speech data - no manual feature engineering necessary, and it trains in hours. In testing, our DNN-based coder performs on par with the AMR -WB standard at a variety of bitrates (~9kbps up to ~24kbps). It also runs in realtime on a 3.8GhZ Intel CPU.

JPEG Image Encryption With Improved Format Compatibility and File Size Preservation

Abstract: Image encryption techniques can be used to ensure the security and privacy of valuable images. The related works in this field have focused more on raster images than on compressed images. Many existing JPEG image encryption schemes are not quite well compatible with the JPEG standard, or the file size of an encrypted JPEG image is apparently increased. In this paper, a novel bitstream-based JPEG image encryption method is presented. First, the groups of successive DC codes that encode the quantized DC coefficient differences with the same sign are permuted within each group. Second, the left half and the right half of a group, whose size will increase with the number of iterations, of consecutive DC codes may be swapped with each other, depending on whether an overflow of quantized DC coefficients occurs during decoding. Third, all AC codes are classified into 63 categories according to their zero-run lengths, then the AC codes within each category are, respectively, scrambled. Finally, all MCUs, except for DC codes, are randomly shuffled as a whole. Moreover, an image-content-related encryption key is employed to provide further security. The experimental results show that the file size of an encrypted JPEG image is almost the same as that of the corresponding plaintext image except for slight variations because of byte alignment. In addition, the quantized DC coefficients decoded from an encrypted JPEG image will not fall outside the valid range. Improved format compatibility is provided compared with other related methods. Moreover, it is unnecessary to perform entropy encoding again because all of the encryption operations are performed directly on the JPEG bitstream. The proposed method proves to be secure against brute-force attacks, differential cryptanalysis, known plaintext attacks, and outline attacks. Our proposed method can also be applied to color JPEG images.

Zero-Delay Rate Distortion via Filtering for Vector-Valued Gaussian Sources

Abstract: We deal with zero-delay source coding of a vector-valued Gauss–Markov source subject to a mean-squared error (MSE) fidelity criterion characterized by the operational zero-delay vector-valued Gaussian rate distortion function (RDF). We address this problem by considering the nonanticipative RDF (NRDF), which is a lower bound to the causal optimal performance theoretically attainable function (or simply causal RDF) and operational zero-delay RDF. We recall the realization that corresponds to the optimal “test-channel” of the Gaussian NRDF, when considering a vector Gauss–Markov source subject to a MSE distortion in the finite time horizon. Then, we introduce sufficient conditions to show existence of solution for this problem in the infinite time horizon (or asymptotic regime). For the asymptotic regime, we use the asymptotic characterization of the Gaussian NRDF to provide a new equivalent realization scheme with feedback, which is characterized by a resource allocation (reverse-waterfilling) problem across the dimension of the vector source. We leverage the new realization to derive a predictive coding scheme via lattice quantization with subtractive dither and joint memoryless entropy coding. This coding scheme offers an upper bound to the operational zero-delay vector-valued Gaussian RDF. When we use scalar quantization, then for $r$ active dimensions of the vector Gauss–Markov source the gap between the obtained lower and theoretical upper bounds is less than or equal to $0.254r + 1$ bits/vector. However, we further show that it is possible when we use vector quantization, and assume infinite dimensional Gauss–Markov sources to make the previous gap to be negligible, i.e., Gaussian NRDF approximates the operational zero-delay Gaussian RDF. We also extend our results to vector-valued Gaussian sources of any finite memory under mild conditions. Our theoretical framework is demonstrated with illustrative numerical experiments.

Hybrid Point Cloud Attribute Compression Using Slice-based Layered Structure and Block-based Intra Prediction

Abstract: Point cloud compression is a key enabler for the emerging applications of immersive visual communication, autonomous driving and smart cities, etc. In this paper, we propose a hybrid point cloud attribute compression scheme built on an original layered data structure. First, a slice partition scheme and a geometry-adaptive k-dimensional tree (k-d tree) method are devised to generate layer structures. Second, we introduce an efficient block-based intra prediction scheme containing to exploit spatial correlations among adjacent points. Third, an adaptive transform scheme based on Graph Fourier Transform (GFT) is Lagrangian optimized to achieve better transform efficiency. The Lagrange multiplier is off-line derived based on the statistics of attribute coding. Last but not least, multiple scan modes are dedicated to improve coding efficiency for entropy coding. Experimental results demonstrate that our method performs better than the state-of-the-art region-adaptive hierarchical transform (RAHT) system, and on average a 37.21% BD-rate gain is achieved. Comparing with the test model for category 1 (TMC1) anchors, which were recently published by MPEG-3DG group on 121st MPEG meeting, a 8.81% BD-rate gain is obtained.

Index Compression Using Byte-Aligned ANS Coding and Two-Dimensional Contexts

Abstract: We examine approaches used for block-based inverted index compression, such as the OptPFOR mechanism, in which fixed-length blocks of postings data are compressed independently of each other. Building on previous work in which asymmetric numeral systems (ANS) entropy coding is used to represent each block, we explore a number of enhancements: (i) the use of two-dimensional conditioning contexts, with two aggregate parameters used in each block to categorize the distribution of symbol values that underlies the ANS approach, rather than just one; (ii) the use of a byte-friendly strategic mapping from symbols to ANS codeword buckets; and (iii) the use of a context merging process to combine similar probability distributions. Collectively, these improvements yield superior compression for index data, outperforming the reference point set by the Interp mechanism, and hence representing a significant step forward. We describe experiments using the 426 GiB gov2 collection and a new large collection of publicly-available news articles to demonstrate that claim, and provide query evaluation throughput rates compared to other block-based mechanisms.

Gaussian Approximation-Based Lossless Compression of Smart Meter Readings

Abstract: Automation metering services, load forecasting, and energy feedback are among the great benefits of smart meters. These meters are usually connected using Narrowband power line communication to transmit the collected waveform readings. The huge volume of these streams, the limited-bandwidth, energy, and required storage space pose a unique management challenge. Compression of these streams has a significant opportunity to solve these issues. Therefore, this paper proposes a new lossless smart meter readings compression algorithm. The uniqueness is in representing smart meter streams using few parameters. This is effectively achieved using Gaussian approximation based on dynamic-nonlinear learning technique. The margin space between the approximated and the actual readings is measured. The significance is that the compression will be only for margin space limited points rather than the entire stream of readings. The margin space values are then encoded using burrow-wheeler transform followed by move-to-front and run-length to eliminate the redundancy. Entropy encoding is finally applied. Both mathematical and empirical experiments have been thoroughly conducted to prove the significant enhancement of the entropy (i.e., almost reduced by half) and the resultant compression ratio (i.e., 3.8:1) which is higher than any known lossless algorithm in this domain.

Quality Assessment of Deep-Learning-Based Image Compression

Abstract: Image compression standards rely on predictive coding, transform coding, quantization and entropy coding, in order to achieve high compression performance. Very recently, deep generative models have been used to optimize or replace some of these operations, with very promising results. However, so far no systematic and independent study of the coding performance of these algorithms has been carried out. In this paper, for the first time, we conduct a subjective evaluation of two recent deep-learning-based image compression algorithms, comparing them to JPEG 2000 and to the recent BPG image codec based on HEVC Intra. We found that compression approaches based on deep auto-encoders can achieve coding performance higher than JPEG 2000, and sometimes as good as BPG. We also show experimentally that the PSNR metric is to be avoided when evaluating the visual quality of deep-learning-based methods, as their artifacts have different characteristics from those of DCT or wavelet-based codecs. In particular, images compressed at low bitrate appear more natural than JPEG 2000 coded pictures, according to a no-reference naturalness measure. Our study indicates that deep generative models are likely to bring huge innovation into the video coding arena in the coming years.

A selective encryption scheme of CABAC based on video context in high efficiency video coding

Abstract: In this paper, we propose a selective encryption scheme with a chaotic encryption system for the High Efficiency Video Coding (HEVC) standard in which Context Adaptive Binary Arithmetic Coding (CABAC) is the only entropy coder for transform coefficient coding. Our method focuses on the binstrings of truncated rice with a context “p” (TRp) code suffix and kth order Exp-Golomb (k = p + 1) code suffix before Binary Arithmetic Coding (BAC) for the remaining absolute level coding, which is coded by the bypass mode in the entropy coding stage. The probability of symbols does not change and CABAC decoding has no effect after encryption. Several different YUV sequences are used for experimental evaluation of the proposed algorithm. Compared to previous researches, our approach has good protection for video information, which keeps a constant bitrate and format compatibility, and meantime it has a negligible impact on encoding performance.

Rate-Distortion Performance of Lossy Compressed Sensing of Sparse Sources

Abstract: We investigate lossy compressed sensing (CS) of a hidden, or remote, source, where a sensor observes a sparse information source indirectly. The compressed noisy measurements are communicated to the decoder for signal reconstruction with the aim to minimize the mean square error distortion. An analytically tractable lower bound to the remote rate-distortion function (RDF), i.e., the conditional remote RDF, is derived by providing support side information to the encoder and decoder. For this setup, the best encoder separates into an estimation step and a transmission step. A variant of the Blahut–Arimoto algorithm is developed to numerically approximate the remote RDF. Furthermore, a novel entropy coding based quantized CS method is proposed. Numerical results illustrate the main rate-distortion characteristics of the lossy CS, and compare the performance of practical quantized CS methods against the proposed limits.

An Efficient Deep Quantized Compressed Sensing Coding Framework of Natural Images

Abstract: Traditional image compressed sensing (CS) coding frameworks solve an inverse problem that is based on the measurement coding tools (prediction, quantization, entropy coding, etc.) and the optimization based image reconstruction method. These CS coding frameworks face the challenges of improving the coding efficiency at the encoder, while also suffering from high computational complexity at the decoder. In this paper, we move forward a step and propose a novel deep network based CS coding framework of natural images, which consists of three sub-networks: sampling sub-network, offset sub-network and reconstruction sub-network that responsible for sampling, quantization and reconstruction, respectively. By cooperatively utilizing these sub-networks, it can be trained in the form of an end-to-end metric with a proposed rate-distortion optimization loss function. The proposed framework not only improves the coding performance, but also reduces the computational cost of the image reconstruction dramatically. Experimental results on benchmark datasets demonstrate that the proposed method is capable of achieving superior rate-distortion performance against state-of-the-art methods.

Spatially adaptive image compression using a tiled deep network

Abstract: Deep neural networks represent a powerful class of function approximators that can learn to compress and reconstruct images Existing image compression algorithms based on neural networks learn quantized representations with a constant spatial bit rate across each image While entropy coding introduces some spatial variation, traditional codecs have benefited significantly by explicitly adapting the bit rate based on local image complexity and visual saliency This paper introduces an algorithm that combines deep neural networks with quality-sensitive bit rate adaptation using a tiled network We demonstrate the importance of spatial context prediction and show improved quantitative (PSNR) and qualitative (subjective rater assessment) results compared to a non-adaptive baseline and a recently published image compression model based on fully-convolutional neural networks

Convolutional Neural Network-Based Arithmetic Coding of DC Coefficients for HEVC Intra Coding

Abstract: In the state-of-the-art video coding standard-High Efficiency Video Coding (HEVC), context-adaptive binary arithmetic coding (CABAC) is adopted as the entropy coding tool. In CABAC, the binarization processes are manually designed, and the context models are empirically crafted, both of which incur that the probability distribution of the syntax elements may not be estimated accurately, and restrict the coding efficiency. In this paper, we adopt a convolutional neural network-based arithmetic coding (CNNAC) strategy, and conduct studies on the coding of the DC coefficients for HEVC intra coding. Instead of manually designing binarization process and context model, we propose to directly estimate the probability distribution of the value of the DC coefficient using densely connected convolutional networks. The estimated probability together with the real DC coefficient are then input into a multi-level arithmetic codec to fulfill entropy coding. Simulation results show that our proposed CNNAC leads to on average 22.47% bits saving compared with CABAC for the bits of DC coefficients, which corresponds to 1.6% BD-rate reduction than the HEVC anchor.

Entropy bounds for grammar compression.

Abstract: Grammar compression represents a string as a context free grammar. Achieving compression requires encoding such grammar as a binary string; there are a few commonly used encodings. We bound the size of practically used encodings for several heuristical compression methods, including \RePair and \Greedy algorithms: the standard encoding of \RePair, which combines entropy coding and special encoding of a grammar, achieves $1.5|S|H_k(S)$, where $H_k(S)$ is $k$-th order entropy of $S$. We also show that by stopping after some iteration we can achieve $|S|H_k(S)$. This is particularly interesting, as it explains a phenomenon observed in practice: introducing too many nonterminals causes the bit-size to grow. We generalize our approach to other compression methods like \Greedy and a wide class of irreducible grammars as well as to other practically used bit encodings (including naive, which uses fixed-length codes). Our approach not only proves the bounds but also partially explains why \Greedy and \RePair are much better in practice than other grammar based methods. In some cases we argue that our estimations are optimal. The tools used in our analysis are of independent interest: we prove the new, optimal, bounds on the zeroth order entropy of parsing of a string.

DCT-Based Color Image Compression Algorithm Using an Efficient Lossless Encoder

Abstract: To enhance the compression ratio of color still image compression, this paper proposes an efficient lossy image compression algorithm using a new efficient lossless encoder. Firstly, the pre-processing, including mean removing and YCbCr transform, is applied to image. Then, this paper applies discrete cosine transform (DCT) to reduce spatial correlation and concentrate the energy of the image. An iterative process based on the bisection method is used to determine the required threshold and control compression quality via achieving the prefixed peak signal-to-noise ratio (PSNR). The next step is applying adaptive scanning to each transform coefficient block to get better compression performance. The final step is the application of a modified lossless encoder to optimize the compression algorithm according to the statistical characteristics of the DCT coefficients. The format of modified encoder is suitable for entropy encoding. Compared with other two algorithms, the experimental results show that the proposed algorithm has better performance in terms of subjective and objective evaluation.

High-Throughput Binary Arithmetic Encoder using Multiple-Bypass Bins Processing for HEVC CABAC

Abstract: The advance of massive video processing applications, devices and resolutions has led to new challenges in video encoding. The HEVC (High Efficiency Video Coding) standard emerges as one alternative in order to address the new video processing requirements. The HEVC allows only one type of entropy encoding algorithm, which is the CABAC (Context-Adaptive Binary Arithmetic Coding). The compression gains achieved by CABAC algorithm come at the cost of increasing complexity for implementation, due to intense data dependencies. The BAE (Binary Arithmetic Encoder) is the CABAC critical sub-block, in which the main part of the algorithm is executed. The present work proposes an 8-stage pipeline BAE architectural solution, named MB-BAE, with the addition of multiple-bypass bins processing, in order to increase throughput without compromising the critical path of the architecture. As a result, an average of 4.94 bins/cycle and around 2.6-Gbin/s of throughput are achieved in our BAE. This is the highest throughput found among related works in the literature, and being able to process 8K UHD videos with the lowest frequency when compared to the same related works.

S-EMG Signal Compression in One-Dimensional and Two-Dimensional Approaches

Abstract: This paper presents algorithms designed for one-dimensional (1-D) and 2-D surface electromyographic (S-EMG) signal compression. The 1-D approach is a wavelet transform based encoder applied to isometric and dynamic S-EMG signals. An adaptive estimation of the spectral shape is used to carry out dynamic bit allocation for vector quantization of transformed coefficients. Thus, an entropy coding is applied to minimize redundancy in quantized coefficient vector and to pack the data. In the 2-D approach algorithm, the isometric or dynamic S-EMG signal is properly segmented and arranged to build a 2-D representation. The high efficient video codec is used to encode the signal, using 16-bit-depth precision, all possible coding/prediction unit sizes, and all intra-coding modes. The encoders are evaluated with objective metrics, and a real signal data bank is used. Furthermore, performance comparisons are also shown in this paper, where the proposed methods have outperformed other efficient encoders reported in the literature.

Edge-Based and Prediction-Based Transformations for Lossless Image Compression

Abstract: Pixelated images are used to transmit data between computing devices that have cameras and screens. Significant compression of pixelated images has been achieved by an “edge-based transformation and entropy coding” (ETEC) algorithm recently proposed by the authors of this paper. The study of ETEC is extended in this paper with a comprehensive performance evaluation. Furthermore, a novel algorithm termed “prediction-based transformation and entropy coding” (PTEC) is proposed in this paper for pixelated images. In the first stage of the PTEC method, the image is divided hierarchically to predict the current pixel using neighboring pixels. In the second stage, the prediction errors are used to form two matrices, where one matrix contains the absolute error value and the other contains the polarity of the prediction error. Finally, entropy coding is applied to the generated matrices. This paper also compares the novel ETEC and PTEC schemes with the existing lossless compression techniques: “joint photographic experts group lossless” (JPEG-LS), “set partitioning in hierarchical trees” (SPIHT) and “differential pulse code modulation” (DPCM). Our results show that, for pixelated images, the new ETEC and PTEC algorithms provide better compression than other schemes. Results also show that PTEC has a lower compression ratio but better computation time than ETEC. Furthermore, when both compression ratio and computation time are taken into consideration, PTEC is more suitable than ETEC for compressing pixelated as well as non-pixelated images.

Device and method for entropy encoding and decoding

Abstract: An entropy encoding device and method, and an entropy decoding device and method are disclosed. The entropy encoding method comprises the steps of: dividing a transform unit into a plurality of zones, and dividing each of the plurality of zones into sub zones; setting, as the first value, the value of the last zone flag, in which a sub zone including a valid transform coefficient among the plurality of zones is included, and setting, as the second value, the value of a zone flag of the remaining zones; setting, as the first value, the value of a sub zone flag of the divided sub zones having the valid transform coefficient, and setting, as the second value, a value of a sub zone flag of the divided sub zones which do not include the valid transform coefficient; determining a preset coefficient coding scheme among a plurality of coefficient coding schemes on the basis of the zone flag and the sub zone flag; encoding a coefficient included in a sub zone on the basis of the determined coefficient coding scheme; and transmitting data on the zone flag, the sub zone flag, and the encoded coefficient.

Entropy Coding and Entropy Coding Improvements of JPEG XS

Abstract: JPEG XS is a new standard for low-latency and low-complexity coding designed by the JPEG committee. Unlike former developments, optimal rate distortion performance is only a secondary goal; the focus of JPEG~XS is to enable cost-efficient, easy to parallelize implementations suitable for FPGAs or GPUs. In this article, we shed some light on the entropy coding back-end of JPEG~XS and introduce modifications of the entropy coding stage currently under discussion that improve objective and subjective quality of the compressed images without compromising the parallelism of the original algorithm.

Improved Efficiency on Adaptive Arithmetic Coding for Data Compression Using Range-Adjusting Scheme, Increasingly Adjusting Step, and Mutual-Learning Scheme

Abstract: Context-based adaptive arithmetic coding (CAAC) has high coding efficiency and is adopted by the majority of advanced compression algorithms. In this paper, five new techniques are proposed to further improve the performance of CAAC. They make the frequency table (the table used to estimate the probability distribution of data according to the past input) of CAAC converge to the true probability distribution rapidly and hence improve the coding efficiency. Instead of varying only one entry of the frequency table, the proposed range-adjusting scheme adjusts the entries near to the current input value together. With the proposed mutual-learning scheme, the frequency tables of the contexts highly correlated to the current context are also adjusted. The proposed increasingly adjusting step scheme applies a greater adjusting step for recent data. The proposed adaptive initialization scheme uses a proper model to initialize the frequency table. Moreover, a local frequency table is generated according to local information. We perform several simulations on edge-directed prediction-based lossless image compression, coefficient encoding in JPEG, bit plane coding in JPEG 2000, and motion vector residue coding in video compression. All simulations confirm that the proposed techniques can reduce the bit rate and are beneficial for data compression.

Entropy Coding Aided Adaptive Subcarrier-Index Modulated OFDM

Abstract: We propose entropy coding-aided adaptive subcarrier-index-modulated orthogonal frequency division multiplexing (SIM-OFDM). In conventional SIM-OFDM, the indices of the subcarriers activated are capable of conveying extra information. We propose the novel concept of compressing the index information bits by employing the Huffman coding. The probabilities of the different subcarrier activation patterns are obtained from an optimization procedure, which improves the performance of the scheme. Both the maximum-likelihood as well as the logarithmic-likelihood ratio-based soft detector may be employed for detecting the subcarriers activated as well as the information mapped to the classic constellation symbols. As an additional advantage of employing the variable-length Huffman codebook, all the legitimate subcarrier activation patterns may be employed, whereas the conventional SIM-OFDM is capable of using only a subset of the patterns. Our simulation results show that an improved performance is attainable by the proposed system.

Wavefront Parallel Processing for AV1 Encoder

Abstract: The emerging AV1 coding standard brings even higher computational complexity than current coding standards, but does not support traditional Wavefront Parallel Processing (WPP) approach due to the lacking of syntax support. In this paper we introduced a novel framework to implement WPP for AV1 encoder that is compatible with current decoder without additional bitstream syntax support, where mode selection is processed in wavefront parallel before entropy encoding and entropy contexts for rate-distortion optimization are predicted. Based on this framework, context prediction algorithms that use same data dependency model as previous works in H.264 and HEVC are implemented. Furthermore, we proposed an optimal context prediction algorithm specifically for AV1. Experimental results showed that our framework with proposed optimal algorithm yields good parallelism and scalability (over 10x speed-up with 16 threads for 4k sequences) with little coding performance loss (less than 0.2% bitrate increasing).

Hybrid Point Cloud Attribute Compression Using Slice-based Layered Structure and Block-based Intra Prediction

Abstract: Point cloud compression is a key enabler for the emerging applications of immersive visual communication, autonomous driving and smart cities, etc. In this paper, we propose a hybrid point cloud attribute compression scheme built on an original layered data structure. First, a slice-partition scheme and geometry-adaptive k dimensional-tree (kd-tree) method are devised to generate the four-layer structure. Second, we introduce an efficient block-based intra prediction scheme containing a DC prediction mode and several angular modes, in order to exploit the spatial correlation between adjacent points. Third, an adaptive transform scheme based on Graph Fourier Transform (GFT) is Lagrangian optimized to achieve better transform efficiency. The Lagrange multiplier is off-line derived based on the statistics of color attribute coding. Last but not least, multiple reordering scan modes are dedicated to improve coding efficiency for entropy coding. In intra-frame compression of point cloud color attributes, results demonstrate that our method performs better than the state-of-the-art region-adaptive hierarchical transform (RAHT) system, and on average a 29.37$\%$ BD-rate gain is achieved. Comparing with the test model for category 1 (TMC1) anchor's coding results, which were recently published by MPEG-3DG group on 121st meeting, a 16.37$\%$ BD-rate gain is obtained.

Accelerating Convolutional Neural Networks via Activation Map Compression

Abstract: The deep learning revolution brought us an extensive array of neural network architectures that achieve state-of-the-art performance in a wide variety of Computer Vision tasks including among others, classification, detection and segmentation. In parallel, we have also been observing an unprecedented demand in computational and memory requirements, rendering the efficient use of neural networks in low-powered devices virtually unattainable. Towards this end, we propose a three-stage compression and acceleration pipeline that sparsifies, quantizes and entropy encodes activation maps of Convolutional Neural Networks. Sparsification increases the representational power of activation maps leading to both acceleration of inference and higher model accuracy. Inception-V3 and MobileNet-V1 can be accelerated by as much as $1.6\times$ with an increase in accuracy of $0.38\%$ and $0.54\%$ on the ImageNet and CIFAR-10 datasets respectively. Quantizing and entropy coding the sparser activation maps lead to higher compression over the baseline, reducing the memory cost of the network execution. Inception-V3 and MobileNet-V1 activation maps, quantized to $16$ bits, are compressed by as much as $6\times$ with an increase in accuracy of $0.36\%$ and $0.55\%$ respectively.

The New CCSDS Standard for Low-Complexity Lossless and Near-Lossless Multispectral and Hyperspectral Image Compression

Abstract: This paper describes the emerging Issue 2 of the CCSDS-123.0-B standard for low-complexity compression of multispectral and hyperspectral imagery, focusing on its new features and capabilities. Most significantly, this new issue incorporates a closed-loop quantization scheme to provide near-lossless compression capability while still supporting lossless compression, and introduces a new entropy coding option that provides better compression of low-entropy data.

Point cloud attribute compression method based on deleting 0 elements in quantisation matrix

Abstract: The invention discloses a point cloud attribute compression method based on deletion of 0 elements in a quantization matrix In view of the quantization matrix during the point cloud attribute compression process, the optimal traversal order is adopted at a coding end to enable 0 elements to be centrally distributed at a tail end in a generated data stream, entropy coding is carried out after the0 are deleted, the data volume of the data stream is reduced, a bit stream generated after coding is reduced, point cloud geometry information is combined at a decoding end to recover the deleted 0 elements, and the method is ensured not to introduce extra errors The method comprises steps: the traversal order of the quantization matrix is optimized at the coding end; the 0 elements at the tail end of the data stream are deleted; the geometry information is consulted to recover the quantization matrix at the decoding end; and a point cloud attribute compression coding process and a decoding process are carried out At the point cloud attribute compression coding end, seven traversal orders are adopted for the quantization matrix, and distribution of the 0 elements in the data stream is more centralized at the tail end; the 0 elements at the tail end of the data stream are deleted, redundant information is removed, and the data volume in need of entropy coding is reduced; and at the decoding end, the point cloud geometry information is combined to complete the deleted 0 elements, the quantization matrix is recovered according to the traversal order, and the compression performanceis improved on the premise of not introducing new errors

Weather Radar Data Compression Based on Spatial and Temporal Prediction

Abstract: The transmission and storage of weather radar products will be an important problem for future weather radar applications. The aim of this work is to provide a solution for real-time transmission of weather radar data and efficient data storage. By upgrading the capabilities of radar, the amount of data that can be processed continues to increase. Weather radar compression is necessary to reduce the amount of data for transmission and archiving. The characteristics of weather radar data are not considered in general-purpose compression programs. The sparsity and data redundancy of weather radar data are analyzed. A lossless compression of weather radar data based on prediction coding is presented, which is called spatial and temporal prediction compression (STPC). The spatial and temporal correlations in weather radar data are utilized to improve the compression ratio. A specific prediction scheme for weather radar data is given, while the residual data and motion vectors are used to replace the original values for entropy coding. After this, the Level-II product from CINRAD SA is used to evaluate STPC. Experimental results show that the STPC achieves a better performance than the general-purpose compression programs, with the STPC yield being approximately 26% better than the next best approach.