Low Power Approaches for Image Acquisition Systems

Abstract

This work presents novel image acquisition methodology to improve power and performance metrics of image acquisition system. Given the slowing Moore’s law, ubiquitous mobile devices like smartphones and focus on multimedia content in today’s world, it is the need of hour to adopt an algorithmic approach to achieve system efficiency in imaging systems. Towards this end, this work employs Compressed Sensing and Deep Learning techniques and tries to find a balance between performance and practicality of implementation. It makes necessary modifications of the algorithms to reduce the entire system redesign efforts which happen to be both expensive and time-consuming process. By following the methodology and trade-offs suggested in this work, one can improve power and performance metrics by 50% while maintaining good quality of final images.