scispace - formally typeset
Search or ask a question
Topic

Quantization (image processing)

About: Quantization (image processing) is a research topic. Over the lifetime, 7977 publications have been published within this topic receiving 126632 citations.


Papers
More filters
Journal ArticleDOI
TL;DR: The proposed method makes use of the interpolated discrete Fourier transform algorithm for estimating the parameters of the fundamental and harmonic components and separates them from the transient ones in the original PQ signal.
Abstract: This paper introduces a high efficient compression method for power quality (PQ) signals. The proposed method makes use of the interpolated discrete Fourier transform algorithm for estimating the parameters (amplitude, frequency, and phase) of the fundamental and harmonic components and separates them from the transient ones in the original PQ signal. When these transient components are submitted to the compression technique, the sparse representation property of the wavelet transform (WT) provides an improvement in the compression ratio (CR). The experimental results show that the proposed method is suitable for various types of PQ signals and that it achieves a higher CR compared to the traditional WT-based compression techniques.

34 citations

Patent
Jungwoo Lee1
28 Sep 1999
TL;DR: In this paper, the authors proposed a parameterized Q matrix adaptation algorithm for MPEG-2 compression, where the Q matrix for the current frame is generated based on DCT coefficient data from the previous encoded frame of the same type (e.g., I, P, or B).
Abstract: In a video compression processing, such as MPEG-2 compression processing, the quantization (Q) matrix used to quantize discrete cosine transform (DCT) coefficients is updated from frame to frame based on a parameterized Q matrix adaptation algorithm. According to the algorithm, the Q matrix for the current frame is generated based on DCT coefficient data (108) from the previous encoded frame of the same type (e.g., I, P, or B) as the current frame. In particular, the Q matrix is generated using a function based on shape parameters (e.g., the slope of the diagonal of the Q matrix and/or the convexity of the diagonal of the Q matrix), where the diagonal slope for the Q matrix of the current frame is generated based on the diagonal slope of a DCT map (106) for the previously encoded frame. Before using the generated Q matrix to quantize the DCT coefficients for the current frame, the Q matrix is preferably adjusted for changes in the target mean from the previously encoded frame to the current frame.

34 citations

Patent
30 Oct 1992
TL;DR: In this article, a quantization parameter for use in encoding a region of an image is developed from a categorization of the region into one of a predetermined plurality of perceptual noise sensitivity (PNS) classes, a level of psycho-visual quality that can be achieved for the encoded version of the image, the level being selected from among a plurality of predetermined levels, and a prestored empirically derived model of the relationship between the PNS classes, the psychovisual quality levels and the values of the quantization parameters.
Abstract: A quantization parameter for use in encoding a region of an image is developed from a) a categorization of the region into one of a predetermined plurality of perceptual noise sensitivity (PNS) classes, b) a level of psycho-visual quality that can be achieved for the encoded version of the image, the level being selected from among a plurality of predetermined levels, and c) a prestored empirically derived model of the relationship between the PNS classes, the psycho-visual quality levels and the values of the quantization parameter. PNS indicates the amount of noise that would be tolerable to a viewer of the region, i.e., the perceptual sensitivity of the region to noise. Some characteristics on which PNS classes may be based are : spatial activity, speed of motion, brightness of the region, importance of the region in a particular context, the presence of edges within the region and the texture of the region, e.g., from "flat" to "highly textured". PNS classes that include combinations of the characteristics of a region of the image may also be defined. The PNS classes employed are selected by the implementor and may be determined empirically. The psycho-visual quality of an encoded image is the quality, as perceived by a viewer, of the version of the image that is reconstructed from the encoded image. It is determined from the complexity of the image and the bit-rate available to encode the image.

34 citations

Proceedings ArticleDOI
22 Jun 2015
TL;DR: This work proposes a temporal order-preserving dynamic quantization method to extract the most discriminative patterns of the action sequence and presents a multimodal feature fusion method that can be derived in this dynamic quantification framework to exploit different discrim inative capability of features from multiple modalities.
Abstract: Recent commodity depth cameras have been widely used in the applications of video games, business, surveillance and have dramatically changed the way of human-computer interaction. They provide rich multimodal information that can be used to interpret the human-centric environment. However, it is still of great challenge to model the temporal dynamics of the human actions and great potential can be exploited to further enhance the retrieval accuracy by adequately modeling the patterns of these actions. To address this challenge, we propose a temporal order-preserving dynamic quantization method to extract the most discriminative patterns of the action sequence. We further present a multimodal feature fusion method that can be derived in this dynamic quantization framework to exploit different discriminative capability of features from multiple modalities. Experiments based on three public human action datasets show that the proposed technique has achieved state-of-the-art performance.

34 citations

Posted Content
03 Apr 2017
TL;DR: This work presents a new approach to learn compressible representations in deep architectures with an end-to-end training strategy based on a soft (continuous) relaxation of quantization and entropy, which is anneal to their discrete counterparts throughout training.
Abstract: In this work we present a new approach to learn compressible representations in deep architectures with an end-to-end training strategy. Our method is based on a soft relaxation of quantization and entropy, which we anneal to their discrete counterparts throughout training. We showcase this method for two challenging applications: Image compression and neural network compression. While these tasks have typically been approached with different methods, our soft-to-hard quantization approach gives state-of-the-art results for both.

34 citations


Network Information
Related Topics (5)
Feature extraction
111.8K papers, 2.1M citations
84% related
Image segmentation
79.6K papers, 1.8M citations
84% related
Feature (computer vision)
128.2K papers, 1.7M citations
84% related
Image processing
229.9K papers, 3.5M citations
83% related
Robustness (computer science)
94.7K papers, 1.6M citations
81% related
Performance
Metrics
No. of papers in the topic in previous years
YearPapers
20228
2021354
2020283
2019294
2018259
2017295