Loading...
Please wait, while we are loading the content...
Similar Documents
Significance-Driven Adaptive Approximate Computing for Energy-E icient Image Processing Applications Special Session Paper : Extended Abstract
| Content Provider | Semantic Scholar |
|---|---|
| Author | Burke, Dave Jenkus, Dainius Qiqieh, Issa Das, Shidhartha Yakovlev, Alex |
| Copyright Year | 2017 |
| Abstract | 1 OVERVIEW With increasing resolutions the volume of data generated by image processing applications is escalating dramatically. When coupled with real-time performance requirements, reducing energy consumption for such a large volume of data is proving challenging. In this paper, we propose a novel approach for image processing applications using signi cance-driven approximate computing. Core to our approach is the fundamental tenet that image data should be processed intelligently based on their informational value, i.e. signi cance. Using quanti ed de nition of signi cance, for the rst time, we show how the complexity of data processing tasks can be drastically reduced when computing decisions are synergistically adapted to signi cance learning principles. A variable-kernel convolution lter case study running on an Odroid XU-4 platform is demonstrated to evaluate the e ectiveness of our approach, with up to 45% energy reduction for a given performance requirement. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://www.staff.ncl.ac.uk/rishad.shafik/files/2015/11/esweek2017acmExtAbs.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
