Simulated Annealing for JPEG Quantization

Hopkins, Max; Mitzenmacher, Michael; Wagner-Carena, Sebastian

doi:10.1109/dcc.2018.00065

Cited by 14 publications

(9 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…achieving a better compression rate than JPEG without losing needed features. Although the quantization table redesign has been proved to be a feasible solution in various applications, such as feature detection [21], visual search [22], it is an intractable optimization problem for "DeepN-JPEG" because of the complexity of parameter searching [23], and the difficulty of a quantitative measurement suitable to DNNs. For example, it is non-trivial to characterize the implicit relationship between image feature (or quantization) errors and DNN accuracy loss.…”

Section: Dnn-oriented Deepn-jpeg Frameworkmentioning

confidence: 99%

DeepN-JPEG: A Deep Neural Network Favorable JPEG-based Image Compression Framework

Liu

Wen

et al. 2018

2018 55th ACM/ESDA/IEEE Design Automation Conference (DAC)

View full text Add to dashboard Cite

As one of most fascinating machine learning techniques, deep neural network (DNN) has demonstrated excellent performance in various intelligent tasks such as image classification. DNN achieves such performance, to a large extent, by performing expensive trainings over huge volumes of training data. To reduce the data storage and transfer overhead in smart resource-limited Internet-of-Thing (IoT) systems, effective data compression is a "must-have" feature before transferring real-time produced dataset for training or classification. While there have been many well-known image compression approaches (such as JPEG), we for the first time find that a human-visual based impage compression approach such as JPEG compression is not an optimized solution for DNN systems, especially with high compression ratios. To this end, we develop an image compression framework tailored for DNN applications, named "DeepN-JPEG", to embrace the nature of deep cascaded information process mechanism of DNN architecture. Extensive experiments, based on "ImageNet" dataset with various state-ofthe-art DNNs, show that "DeepN-JPEG" can achieve ∼ 3.5× higher compression rate over the popular JPEG solution while maintaining the same accuracy level for image recognition, demonstrating its great potential of storage and power efficiency in DNN-based smart IoT system design.

show abstract

Section: Dnn-oriented Deepn-jpeg Frameworkmentioning

confidence: 99%

DeepN-JPEG: A Deep Neural Network Favorable JPEG-based Image Compression Framework

Liu

Wen

et al. 2018

2018 55th ACM/ESDA/IEEE Design Automation Conference (DAC)

View full text Add to dashboard Cite

show abstract

“…Y, where X and Y are image dimensions. In addition, Structural Similarity Index (SSIM) is used as a subjective quality measurement for the test images besides the PSNR, SSIM value ranges between 0.0-1.0, where low value means large structural variation, and vice versa [11,22]. Four tests are carried out to show the improvements of the proposed method compared to JPEG compression.…”

Section: Resultsmentioning

confidence: 99%

“…Singh proposed a deblocking algorithm for filtering those blocked boundaries by making use of smoothening, detection of blocked edges, and filtering only the difference between the pixels that contain the blocked edge [2]. Finally, Hopkins et al improved JPEG compression quality through searching for new quantization tables that have the ability to decrease the FSIM (feature Similarity Index Measure) error and increase CR (Compression Ratio) at certain quality levels [22].…”

Section: Introductionmentioning

confidence: 99%

Pre and Postprocessing for JPEG to Handle Large Monochrome Images

2019

View full text Add to dashboard Cite

Image compression is one of the most important fields of image processing. Because of the rapid development of image acquisition which will increase the image size, and in turn requires bigger storage space. JPEG has been considered as the most famous and applicable algorithm for image compression; however, it has shortfalls for some image types. Hence, new techniques are required to improve the quality of reconstructed images as well as to increase the compression ratio. The work in this paper introduces a scheme to enhance the JPEG algorithm. The proposed scheme is a new method which shrinks and stretches images using a smooth filter. In order to remove the blurring artifact which would be developed from shrinking and stretching the image, a hyperbolic function (tanh) is used to enhance the quality of the reconstructed image. Furthermore, the new approach achieves higher compression ratio for the same image quality, and/or better image quality for the same compression ratio than ordinary JPEG with respect to large size and more complex content images. However, it is an application for optimization to enhance the quality (PSNR and SSIM), of the reconstructed image and to reduce the size of the compressed image, especially for large size images.

show abstract

“…Lastly, the Simulated Annealing algorithm was recently explored as an alternative to design quantization tables in Hopkins et al [22], optimizing over an image dataset, finding static (not image specific) quantization tables that considerably outperforms JPEG standard tables.…”

Section: Related Workmentioning

confidence: 99%

Nature Inspired JPEG Quantization Optimization

Ferreira¹,

Peixoto²,

Júnior³

2020

Anais De XXXVIII Simpósio Brasileiro De Telecomunicações E Processamento De Sinais

View full text Add to dashboard Cite

The JPEG standard is the most used algorithm for image compression, with billions of decoders deployed in either software or hardware. The standard allows for custom quantization tables, that are transmitted in the bitstream. This work examines the application of two promising nature inspired metaheuristics, Particle Swarm Optimization and Dual Simulated Annealing, in the generation of custom, image-specific, ratedistortion optimal quantization tables for the JPEG. Our results show that this approach is capable of producing a JPEG compliant, backwards compatible images that presents a compression rate about 9% smaller than standard JPEG.

show abstract

Simulated Annealing for JPEG Quantization

Cited by 14 publications

References 10 publications

DeepN-JPEG: A Deep Neural Network Favorable JPEG-based Image Compression Framework

DeepN-JPEG: A Deep Neural Network Favorable JPEG-based Image Compression Framework

Pre and Postprocessing for JPEG to Handle Large Monochrome Images

Nature Inspired JPEG Quantization Optimization

Contact Info

Product

Resources

About