[month] [year]

Shivani Maurya – Image Processing Algorithms

Shivani Maurya received her Master of Science – Dual Degree in  Electronics and Communication Engineering (ECE).  Her research work was supervised by Dr. Suresh Purini. Here’s a summary of her research work on Accuracy Configurable FPGA Implementation of Image Processing Algorithms:

Image processing algorithms with intrinsic robustness to errors can be approximated for significant resource and energy savings while still meeting the end-user requirements. FPGA-based implementations can increase their suitability for real-time high-speed multimedia applications by leveraging Approximate Computing, an independent field that explores methods to reduce computation costs by allowing minor degradation in intermediate computations. With high volume of pixel level computations, algorithms like the Harris Corner Detector (HCD) and Unsharp Mask (USM), offer a wide range of targets for such strategies. In this thesis, we propose hardware implementations of HCD and USM that rely on approximating the intermediate multiplication operations using Dynamic Range Unbiased Multiplier (DRUM). With configurable bit-width control of DRUM instances, the visual quality of outputs is shown to depend on the varying accuracy of the architecture. We explore how the errors due to approximate operations propagate to the corner response and pixel classification. We then attempt to find a threshold that adapts to this inaccuracy across images. The proposed approximate architecture utilized less than 50% of the resources of the Xilinx Virtex-7 and Zynq-7000 FPGA platforms. When compared with analogous accurate implementations of HCD and USM on other FPGA devices, the BRAM and DSP usage of our design was found to be the least. Synthesis results show that the proposed implementation achieves over 60% increase in maximum frequency compared to the base implementation. The HCD and USM architectures were further qualified using application specific metrics which were instrumental in determining the permissible bit-widths for the DRUM instances at different pipeline stages to achieve desired output quality. Our approach can also be applied to other signal processing and computer vision applications governed by respective quality metrics.

Keywords – Approximate Computing, DRUM Multiplier, Image Processing Benchmarks, Harris Corner Detection, Unsharp Mask

January 2023