En-hui Yang

Distortion Comparisons On Successive Refinement Over Layered Broadcast Channel With Protection

IEEE Transactions on Multimedia, 2013

Entropy Estimators with Almost Sure Convergence and an O(n-1) Variance

Entropy Estimators with Almost Sure Convergence and an O(n−1) Variance Alexei Kaltchenko Physics ... more

Fixed-slope universal algorithms for lossy source coding via lossless codeword length functions

The main advantage of the fixed slope universal lossy data compression scheme over the fixed rate... more The main advantage of the fixed slope universal lossy data compression scheme over the fixed rate (fixed distortion) universal lossy data compression scheme lies in the fact that it converts the encoding problem to a search problem through a trellis and then permits one to use some sequential search algorithms to implement it. Simulation results with the kth order arithmetic codeword length function as a lossless codeword length function and the M-algorithm as a sequential search algorithm show that this fixed slope universal algorithm, combined with suitable search algorithms, might be implementable in practice.

Low-Density Parity-Check Code with Fast Decoding Speed

arXiv (Cornell University), Feb 23, 2006

Low-Density Parity-Check (LDPC) codes received much attention recently due to their capacity-appr... more Low-Density Parity-Check (LDPC) codes received much attention recently due to their capacity-approaching performance. The iterative message-passing algorithm is a widely adopted decoding algorithm for LDPC codes [7]. An important design issue for LDPC codes is designing codes with fast decoding speed while maintaining capacityapproaching performance. In another words, it is desirable that the code can be successfully decoded in few number of decoding iterations, at the same time, achieves a significant portion of the channel capacity. Despite of its importance, this design issue received little attention so far. In this paper, we address this design issue for the case of binary erasure channel. We prove that density-efficient capacity-approaching LDPC codes satisfy a so called "flatness condition". We show an asymptotic approximation to the number of decoding iterations. Based on these facts, we propose an approximated optimization approach to finding the codes with good decoding speed. We further show that the optimal codes in the sense of decoding speed are "right-concentrated". That is, the degrees of check nodes concentrate around the average right degree.

Download

On Optimal Information-Theoretically Secure Key Management

2019 16th Canadian Workshop on Information Theory (CWIT), 2019

With an underlying secure symmetric cipher of key length l, this paper addresses how best to secu... more

Dual set arithmetic coding and its applications to image coding

1996 8th European Signal Processing Conference (EUSIPCO 1996), 1996

Arithmetic coding is usually implemented in fixed precision. Such an implementation cannot effici... more Arithmetic coding is usually implemented in fixed precision. Such an implementation cannot efficiently code sources, such as image coding algorithms, that locally produce a small fraction of a large alphabet of symbols. In this paper, we propose a novel approach to overcome this inefficiency. The proposed algorithm uses dual symbol sets: a primary symbol set that contains the symbols that have occurred in the recent past and a secondary symbol set that contains all other symbols. Both sets are dynamically adapted to the local statistics. We summarize an analysis of the proposed approach and describe the results that we have obtained by applying it to images.

Download

Targeted Attack for Deep Hashing Based Retrieval

Computer Vision – ECCV 2020, 2020

The deep hashing based retrieval method is widely adopted in large-scale image and video retrieva... more The deep hashing based retrieval method is widely adopted in large-scale image and video retrieval. However, there is little investigation on its security. In this paper, we propose a novel method, dubbed deep hashing targeted attack (DHTA), to study the targeted attack on such retrieval. Specifically, we first formulate the targeted attack as a point-to-set optimization, which minimizes the average distance between the hash code of an adversarial example and those of a set of objects with the target label. Then we design a novel component-voting scheme to obtain an anchor code as the representative of the set of hash codes of objects with the target label, whose optimality guarantee is also theoretically derived. To balance the performance and perceptibility, we propose to minimize the Hamming distance between the hash code of the adversarial example and the anchor code under the ∞ restriction on the perturbation. Extensive experiments verify that DHTA is effective in attacking both deep hashing based image retrieval and video retrieval.

Download

Structural complexity of random binary trees

2009 IEEE International Symposium on Information Theory, 2009

For each positive integer n, let T n be a random rooted binary tree having finitely many vertices... more For each positive integer n, let T n be a random rooted binary tree having finitely many vertices and exactly n leaves. We can view H(T n), the entropy of T n , as a measure of the structural complexity of tree T n in the sense that approximately H(Tn) bits suffice to construct Tn. We are interested in determining conditions on the sequence (T n : n = 1, 2, • • •) under which H(T n)/n converges to a limit as n → ∞. We exhibit some of our progress on the way to the solution of this problem.

Download

Universal compression of ergodic quantum sources

Quantum Information & Computation, Jul 1, 2003

1) For a real r > 0, let F (r) be the family of all stationary ergodic quantum sources with von N... more 1) For a real r > 0, let F (r) be the family of all stationary ergodic quantum sources with von Neumann entropy rate less than r. We prove that, for any r > 0, there exists a blind, source-independent block compression scheme which compresses every source from F (r) to rn qubits per input block of size n with arbitrary high fidelity for all large enough n. 2) We show that the stationarity and the ergodicity of a quantum source {ρ m } ∞ m=1 are preserved by any trace-preserving completely positive linear map of the tensor product form E ⊗m , where a copy of E acts locally on each spin lattice site. We also establish ergodicity criteria for so called classicallycorrelated quantum sources.

Download

Rate Distortion Optimization in H . 264

In this paper, we study the best rate distortion performance that an H.264 encoder can possibly a... more In this paper, we study the best rate distortion performance that an H.264 encoder can possibly achieve. Using soft decision quantization rather than the standard hard decision quantization, we first establish a general framework for jointly designing motion compensation, quantization, and entropy coding in the hybrid coding structure of H.264 to minimize a true rate distortion cost. We then propose three rate distortion optimization algorithms—a graph-based algorithm for optimal soft decision quantization in H.264 baseline profile encoding given motion compensation and quantization step sizes, an iterative algorithm for optimal residual coding in H.264 baseline profile encoding given motion compensation, and an iterative overall algorithm for optimal H.264 baseline profile encoding—with them embedded in the indicated order. The graph-based algorithm for optimal soft decision quantization is the core; given motion compensation and quantization step sizes, it is guaranteed to perform...

On the design and analysis of grammar-based data compression algorithms

The grammar-based coding theory is a new development in the field of source coding. It provides a... more The grammar-based coding theory is a new development in the field of source coding. It provides a framework to design lossless data compression algorithms by combining the power of string matching and that of statistical modelling. In this theory, each algorithm first transforms a sequence to be compressed into a grammar, and then uses an arithmetic code to compress the grammar. In this thesis research, we investigate the following topics of the design and analysis of grammar-based data compression algorithms: (1) Arithmetic coding: To meet the special requirement on arithmetic coding posed by the grammar-based coding theory, a new arithmetic coding scheme called the multilevel arithmetic coding is proposed to encode sequences from large and unbounded alphabets. The optimality of the multilevel arithmetic coding scheme is established. A greedy renormalization method is presented to significantly reduce the computational complexity of arithmetic coding. (2) Implementation and complex...

<title>Universal compression of classically correlated stationary quantum sources</title>

Quantum Information and Computation, 2003

For a real number r > 0, let F(r) be the family of all stationary completely ergodic quantum s... more For a real number r > 0, let F(r) be the family of all stationary completely ergodic quantum sources with von Neumann entropy rates less than r. We prove that, for any r > 0, there exists a blind, source-independent block compression scheme which compresses every source from F(r) to rn qubits per input block length n with arbitrarily high fidelity for all large n. As our second result, we show that the stationarity and the ergodicity of a quantum source {ρm}m=1∞ are preserved by any trace-preserving completely positive linear map of the tensor product form ε⊗m, where a copy of ε acts locally on each spin lattice site. We also establish ergodicity criteria for so called classically-correlated quantum sources.

A Watermarking-Based Framework for Protecting Deep Image Classifiers Against Adversarial Attacks

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2021

Although deep learning-based models have achieved tremendous success in image-related tasks, they... more Although deep learning-based models have achieved tremendous success in image-related tasks, they are known to be vulnerable to adversarial examples-inputs with imperceptible, but subtly crafted perturbation which fool the models to produce incorrect outputs. To distinguish adversarial examples from benign images, in this paper, we propose a novel watermarking-based framework for protecting deep image classifiers against adversarial attacks. The proposed framework consists of a watermark encoder, a possible adversary, and a detector followed by a deep image classifier to be protected. Specific methods of watermarking and detection are also presented. It is shown by experiment on a subset of ImageNet validation dataset that the proposed framework along with the presented methods of watermarking and detection is effective against a wide range of advanced attacks (static and adaptive), achieving a near zero (effective) false negative rate for FGSM and PGD attacks (static and adaptive) with the guaranteed zero false positive rate. In addition, for all tested deep image classifiers (ResNet50V2, MobileNetV2, InceptionV3), the impact of watermarking on classification accuracy is insignificant with, on average, 0.63% and 0.49% degradation in top 1 and top 5 accuracy, respectively.

Download

Closed-Form Formulas for Private Watermarking Capacities of Laplacian Sources with the Magnitude-Error Distortion Measure and Under Additive Attacks

Calculation of watermarking capacities of private Laplacian watermarking systems with the magnitu... more Calculation of watermarking capacities of private Laplacian watermarking systems with the magnitude-error distortion measure under fixed attacks is addressed. First, in the case of an additive Laplacian attack, a nice closed-form formula for the watermarking capacities is derived, which involves only the distortion level and the parameter of the Laplacian attack. Second, in the case of an arbitrary additive attack, a general, but slightly more complicated formula for the watermarking capacities is given. Finally, calculation of the joint compression and private watermarking rate region of Laplacian watermarking systems with an additive Laplacian attack is considered.

Download

Fast inter mode decision for HEVC based on transparent composite model

2015 IEEE International Conference on Image Processing (ICIP), 2015

In comparison with H.264/AVC, the newest video coding standard, High Efficiency Video Coding (HEV... more In comparison with H.264/AVC, the newest video coding standard, High Efficiency Video Coding (HEVC), improves video coding rate distortion (RD) performance, but at the price of significant increase in its encoding complexity, due to its complicated inter mode decision process. In HEVC inter coding, the actual RD costs of all combinations of coding units (CUs), prediction units (PUs), and transform units (TUs) have to be computed and then the combination (i.e., mode) with the minimum cost is selected and encoded. To reduce the inter mode decision complexity in HEVC while maintaining its coding efficiency, in this paper, a fast inter mode decision method based on a newly proposed Transparent Composite Model (TCM) is developed. Spatially and temporally homogeneities are identified by TCM, and then used, together with spatiotemporal correlation between CUs, to determine PU and CU partitions so that cost computations of unnecessary modes can be skipped. Experimental results show that, for the low delay main test configuration of HEVC, our method reduces, on average, the encoding time by 60.71% with an insignificant loss in coding efficiency (1.06% BD-Rate increase).

Two results on interactive lossless source encoding and decoding with side information at the decoder

2008 Third International Conference on Communications and Networking in China, 2008

ABSTRACT Slepian and Wolf proposed and analyzed a ground-breaking paradigm of source coding for a... more ABSTRACT Slepian and Wolf proposed and analyzed a ground-breaking paradigm of source coding for a source network in which a finite alphabet source X is to be encoded and transmitted to the decoder, and another finite alphabet source Y correlated with X is available only to the decoder as a helper, or in other words, the side information. In Slepian and Wolf&#39;s paradigm, which is now commonly referred to as Slepian-Wolf (SW) coding, information flow is one-way from the encoder to the decoder, and no interaction is allowed between them.In this paper, we consider a paradigm of source coding called interactive encoding and decoding (IED). The paper answers the question: what can be gained by IED over SW coding?

The compression performance of grammar-based codes revisited

Proceedings IEEE International Symposium on Information Theory,

ABSTRACT A grammar-based code (GC) refers to a new type of universal lossless source code. The au... more ABSTRACT A grammar-based code (GC) refers to a new type of universal lossless source code. The authors consider the following two questions: (1) Based on run-length encoding (RLE) can one propose a class of interesting algorithms which is superior to arithmetic coding (AC) algorithms with finite contexts and can also be used as a benchmark to evaluate GCs? (2) For each sequence x, is there a better quantity than rk*(x) such that it can be used to derive a stronger redundancy bound for GCs? They show that the answer to both questions is yes and summarize the main results.

On the Codebook-Level Duality Between Slepian-Woif Coding and Channel Coding

2007 Information Theory and Applications Workshop, 2007

A codebook-level duality between Slepian-Wolf coding and channel coding is established. Specifica... more A codebook-level duality between Slepian-Wolf coding and channel coding is established. Specifically, it is shown that using linear codes over ZM (the ring of integers mod M), each Slepian-Wolf coding problem is equivalent to a channel coding problem for a semi-symmetric additive channel under optimal decoding, belief propagation decoding, and minimum entropy decoding. Various notions of symmetric channels are discussed and their connections with semi-symmetric additive channels are clarified.

Download

Algorithms for Computing Joint Compression and Private Watermarking Rate Regions

2006 IEEE International Symposium on Information Theory, 2006

Based on the idea of the Blahut-Arimoto algorithm for computing channel capacities and rate-disto... more Based on the idea of the Blahut-Arimoto algorithm for computing channel capacities and rate-distortion functions, two iterative algorithms are developed for numerical computation of the compression and watermarking rate regions of joint compression and private watermarking systems with finite alphabets. The convergence of the algorithms developed is also proved

A simple technique for bounding the pointwise redundancy of the 1978 Lempel-Ziv algorithm

Proceedings of the Data Compression Conference

If x is a string of finite length over a finite alphabet A, let LZ(x) denote the length of the bi... more If x is a string of finite length over a finite alphabet A, let LZ(x) denote the length of the binary codeword assigned to x by the 1978 version of the Lempel-Ziv data compression algorithm, let t(x) be the number of phrases in the Lempel-Ziv parsing of x, and let μ(x) be the probability assigned to x by a memoryless source model. Using a very simple technique, we probe the pointwise redundancy bound LZ(x)+log2μ(x)&les;8t(x)max{-log2 μ(a):a∈A}

Download

Uploads

Papers by En-hui Yang

Log In