Vector quantization and signal compression

By A. Gersho, Robert M. Gray. 1991

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Evaluation & Benchmarks

Citations in the corpus

GAN Inversion: A Survey (2021, methodology): “PCA enforces a weak orthogonality constraint, resulting in a distributed representation that uses cancellations to capture global variability [191], [192], while VQ uses a strong winner-take-all constraint that leads to clustering the data into mutually exclusive prototypes [193].”
End-to-end Optimized Image Compression (2016, background): “Without further constraints, the general problem of optimal quantization in high-dimensional spaces is intractable (Gersho and Gray, 1992).”
Price of Transparency in Strategic Machine Learning (2016, methodology): “We use standard game theoretic notations for the related results throughout this paper (cf. [7]).”
High-quality Speech Coding with Sample RNN (2018, methodology): “combinations of source coding techniques are utilized to achieve encoding efﬁciency with appropriate perceptual consideration, including vector quantization (VQ), predictive coding and entropy coding [18]. In this paper, for all experiments, we deﬁne the operating points of the encoder as in Table 1”
Propagation kernels: efficient graph kernels from propagated information (2014, background): “where pt,u is the node-label distribution of node u at iteration t and h (·) is a quantization function (Gersho and Gray 1991), more precisely a locality sensitive hash (lsh) function (Datar and Indyk 2004), which will be introduced in more detail in the next section.”
Technologies for 3D mesh compression: A survey (2005, background): “Compared with scalar quantization, the main advantages of VQ include a superior R-D performance , more freedom in choosing quantization cell shapes, and better exploitation of dependence between vector components [36].”
Content-Based Image Retrieval Using Multiresolution Color and Texture Features (2008, background): “To understand the characteristics of BDIP more deeply, let us consider two operators which are defined as follows: (4) (5) where denotes the magnitude of representative gradient in an window whose center pixel is and thus the first operator performs gradient operation.”
High-Dimensional Stochastic Gradient Quantization for Communication-Efficient Edge Learning (2019, background): “There are few results on the VQ of gradients despite its being a well-developed area [11].”
Analysis of approximate nearest neighbor searching with clustered point sets (1999, unknown): “Nearest neighbor searching has applications in many areas, including knowledge discovery and data mining [FPSSU96], pattern recognition and classification [CH67, DH73], machine learning [CS93], data compression [GG92], multimedia databases [FSN95], document retrieval [DDF90], and statistics [DW82].”
On the construction of some capacity-approaching coding schemes (2000, methodology): “For high-rate quantization, the pdf of the quantization points should be proportional to p(w) for optimality, where p(w) is the source pdf of W and r is the exponent in the distortion measure ||·||r [29].”
The Effectiveness of Lloyd-Type Methods for the k-Means Problem (2006, background): “This and similar iterative descent methods soon became the dominant approaches to the problem [35, 33, 12, 31] (see also [19, 20, 24] and the references therein); they remain so today, and are still being improved [1, 40, 42, 28].”
Discretizing Unobserved Heterogeneity (2017, methodology): “Given parameter estimates α̂(k, j), ĉ(k, j), σ̂(2), and θ̂, we update the estimated partition of individuals using the full model’s structure, as in (20) and (21); see Appendix H.”
Survey of maneuvering target tracking. Part V. Multiple-model methods (2005, background): “The problem of (scalar or vector) quantization and data compression [121] is in a sense a special case of the model-set design problem in this formulation.”
From Sparse Solutions of Systems of Equations to Sparse Modeling of Signals and Images (2009, methodology): “Furthermore, in such a case the above training algorithms simplify to the well-known K-means algorithm [81].”
Robust and secure image hashing (2006, methodology): “where are key-dependent pseudorandom numbers that are normally distributed with mean and variance . This method is illustrated in Fig. 3. 3) Post Processing: We quantize the resulting statistics vector and apply gray coding to obtain the binary hash sequence [ 37 ].”
Iterative Gaussianization: From ICA to Random Rotations (2011, background): “MANY signal processing problems such as coding, restoration, classification, regression or synthesis greatly depend on an appropriate description of the underlying probability density function (PDF) [1]–[5].”