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: A novel watermarking method based on a discrete cosine transform (DCT) which guarantees robustness and low computational complexity is proposed, which had the faster and the more robust performance than previous studies.
Abstract: In many studies related to watermarking, spatial-domain methods have a relatively low information-hiding capacity and limited robustness, and transform-domain methods are not applicable in real-time processes because of their considerably high computational time. In this paper, we propose a novel watermarking method based on a discrete cosine transform (DCT), which guarantees robustness and low computational complexity. First, we calculated the DCT coefficient of a specific location. Then, a variation value was calculated according to the embedding bits and quantization steps to modify the coefficient. Last, we embedded watermark bits by directly modifying the pixel values without full-frame DCT. Tests comparing invisibility, robustness, and computational time were conducted for determining the feasibility of the proposed method. The results showed that the proposed method had the faster and the more robust performance than previous studies.

31 citations

Posted Content
TL;DR: This paper proposes a framework to jointly prune and quantize the DNNs automatically according to a target model size without using any hyper-parameters to manually set the compression ratio for each layer.
Abstract: Deep Neural Networks (DNNs) are applied in a wide range of usecases. There is an increased demand for deploying DNNs on devices that do not have abundant resources such as memory and computation units. Recently, network compression through a variety of techniques such as pruning and quantization have been proposed to reduce the resource requirement. A key parameter that all existing compression techniques are sensitive to is the compression ratio (e.g., pruning sparsity, quantization bitwidth) of each layer. Traditional solutions treat the compression ratios of each layer as hyper-parameters, and tune them using human heuristic. Recent researchers start using black-box hyper-parameter optimizations, but they will introduce new hyper-parameters and have efficiency issue. In this paper, we propose a framework to jointly prune and quantize the DNNs automatically according to a target model size without using any hyper-parameters to manually set the compression ratio for each layer. In the experiments, we show that our framework can compress the weights data of ResNet-50 to be 836$\times$ smaller without accuracy loss on CIFAR-10, and compress AlexNet to be 205$\times$ smaller without accuracy loss on ImageNet classification.

31 citations

Proceedings ArticleDOI
19 Mar 2018
TL;DR: In this paper, the authors describe the making of a real-time object detection in a live video stream processed on an embedded all-programmable device, where the required processing is tamed and parallelized across both the CPU cores and the programmable logic and how the most suitable resources and powerful extensions, such as NEON vectorization, are leveraged for the individual processing steps.
Abstract: Neural networks have established as a generic and powerful means to approach challenging problems such as image classification, object detection or decision making. Their successful employment foots on an enormous demand of compute. The quantization of network parameters and the processed data has proven a valuable measure to reduce the challenges of network inference so effectively that the feasible scope of applications is expanded even into the embedded domain. This paper describes the making of a real-time object detection in a live video stream processed on an embedded all-programmable device. The presented case illustrates how the required processing is tamed and parallelized across both the CPU cores and the programmable logic and how the most suitable resources and powerful extensions, such as NEON vectorization, are leveraged for the individual processing steps. The crafted result is an extended Darknet framework implementing a fully integrated, end-to-end solution from video capture over object annotation to video output applying neural network inference at different quantization levels running at 16 frames per second on an embedded Zynq UltraScale+ (XCZU3EG) platform.

31 citations

Journal ArticleDOI
TL;DR: In this method, the closest-neighboring pointers are used to enlarge the search range of each search path and guide all search paths into more appropriate codewords, which improves the image quality and the encoding time of tree-structured vector quantization.
Abstract: We propose a new method to improve the image quality and the encoding time of tree-structured vector quantization (TSVQ). We call this new method the closest-coupled tree-structured vector quantization (CCTSVQ). In this method, we use the closest-neighboring pointers to enlarge the search range of each search path and guide all search paths into more appropriate codewords. CCTSVQ therefore improves the image quality of TSVQ. This fact is shown in our experimental results. Furthermore, in these experimental results, we see that the encoding time of CCTSVQ is always faster than that of TSVQ, based on the same image quality.

31 citations

Patent
Jeon Jong-Gu1
19 Oct 1994
TL;DR: In this paper, the scale factor of the quantization step size of the discrete cosine transform coefficients is controlled in a macro block in accordance with a predetermined compression ratio, and summing the complexity normalizing signal, the zero coefficient normalizing value and the transient weight signal so as to produce a scale factor.
Abstract: An image coding method includes the steps of producing a complexity normalizing signal for respective macro blocks based on a complexity in a space domain with regard to an input image, producing a zero coefficient sum by performing a discrete cosine transform operation on the input image and counting the discrete cosine transform coefficients from -1 to +1, producing a zero coefficient normalizing value for respective macro blocks in accordance with the zero coefficient sum; producing a transient weight signal for respective macro blocks in accordance with the zero coefficient sum and a predetermined compression ratio, and summing the complexity normalizing signal, the zero coefficient normalizing value and the transient weight signal so as to produce the scale factor. The quantization step size of the discrete cosine transform coefficients is controlled in a macro block in accordance with the scale factor. Thus, image frames having various complexities can be encoded at a constant bit rate with no deterioration of the image.

31 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