Approximate compression: enhancing compressibility through data approximation

Internet-connected mobile processors used in cellphones, tablets, and internet-of-things (IoT) devices are generating and transmitting data at an ever-increasing rate. These devices are already the most abundant types of processor parts produced and used today and are growing in ubiquity with the rapid proliferation of mobile and IoT technologies. Size and usage characteristics of these data-generating systems dictate that they will continue to be both bandwidth- and energy-constrained. The most popular mobile applications, dominating communication bandwidth utilization for the entire internet, are centered around transmission of image, video, and audio content. For such applications, where perfect data quality is not required, approximate computation has been explored to alleviate system bottlenecks by exploiting implicit noise tolerance to trade off output quality for performance and energy benefits. However, it is often communication, not computation, that dominates performance and energy requirements in mobile systems. This is coupled with the increasing tendency to offload computation to the cloud, making communication efficiency, not computation efficiency, the most critical parameter in mobile systems. Given this increasing need for communication efficiency, data compression provides one effective means of reducing communication costs. In this paper, we explore approximate compression and communication to increase energy efficiency and alleviate bandwidth limitations in communication-centric systems. We focus on application-specific approximate data compression, whereby a transmitted data stream is approximated to improve compression rate and reduce data transmission cost. Whereas conventional lossy compression follows a one-size-fits-all mentality in selecting a compression technique, we show that higher compression rates can be achieved by understanding the characteristics of the input data stream and the application in which it is used. We introduce a suite of data stream approximations that enhance the compression rates of lossless compression algorithms by gracefully and efficiently trading off output quality for increased compression rate. For different classes of images, we explain the interaction between compression rate, output quality, and complexity of approximation and establish comparisons with existing lossy compression algorithms. Our approximate compression techniques increase compression rate and reduce bandwidth utilization by up to 10X with respect to state-of-the-art lossy compression while achieving the same output quality and better end-to-end communication performance.

[1]  Truong Q. Nguyen,et al.  Integer fast Fourier transform , 2002, IEEE Trans. Signal Process..

[2]  Sang Joon Kim,et al.  A Mathematical Theory of Communication , 2006 .

[3]  Mohamed S. Abdelfattah,et al.  Gzip on a chip: high performance lossless data compression on FPGAs using OpenCL , 2014, IWOCL '14.

[4]  Martín Abadi,et al.  TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems , 2016, ArXiv.

[5]  Gregory K. Wallace,et al.  The JPEG still picture compression standard , 1991, CACM.

[6]  F. Keissarian.  A New Quadtree-based Image Compression Technique using Pattern Matching Algorithm , 2009 .

[7]  Eero P. Simoncelli,et al.  Image quality assessment: from error visibility to structural similarity , 2004, IEEE Transactions on Image Processing.

[8]  Touradj Ebrahimi,et al.  The JPEG 2000 still image compression standard , 2001, IEEE Signal Process. Mag..

[9]  David Salomon,et al.  Data Compression: The Complete Reference , 2006 .

[10]  Alvin,et al.  Wallace the JPEG still picture compression standard ' , 2017 .

[11]  Soontorn Oraintara,et al.  Integer fast Fourier transform (INTFFT) , 2001, 2001 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.01CH37221).

[12]  H. Zappe,et al.  Frequency analysis of maskless lithography system for generation of continuous phase patterns with homogeneous structure depth , 2008 .

[13]  Zhou Wang,et al.  Multiscale structural similarity for image quality assessment , 2003, The Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, 2003.

[14]  Willie Anderson,et al.  Hexagon DSP: An Architecture Optimized for Mobile Multimedia and Communications , 2014, IEEE Micro.

[15]  Sergey Ioffe,et al.  Rethinking the Inception Architecture for Computer Vision , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[16]  Renato J. Cintra,et al.  An Integer Approximation Method for Discrete Sinusoidal Transforms , 2020, Circuits Syst. Signal Process..

[17]  Dumitru Erhan,et al.  Going deeper with convolutions , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[18]  Fábio M. Bayer,et al.  A DCT Approximation for Image Compression , 2011, IEEE Signal Processing Letters.

[19]  Dave Evans,et al.  How the Next Evolution of the Internet Is Changing Everything , 2011 .

[20]  Vikram Bhatt,et al.  The GreenDroid Mobile Application Processor: An Architecture for Silicon's Dark Future , 2011, IEEE Micro.

[21]  Joo-Young Kim,et al.  A Scalable High-Bandwidth Architecture for Lossless Compression on FPGAs , 2015, 2015 IEEE 23rd Annual International Symposium on Field-Programmable Custom Computing Machines.

[22]  Antonio Ortega,et al.  Scalable variable complexity approximate forward DCT , 2004, IEEE Transactions on Circuits and Systems for Video Technology.

[23]  Joan L. Mitchell,et al.  JPEG: Still Image Data Compression Standard , 1992 .