Single Instruction Multiple Data and Cellular Non-linear Networks as Fine-Grained Parallel Solutions for Early Vision on FPGAs

This paper examines the feasibility of fine-grained parallelism on FPGAs for early vision. The paper compares the performance and functionality of Cellular Non-linear Networks and Single Instruction Multiple Data architectures on FPGAs. Area and speed data along with different examples of low-level image processing tasks are discussed throughout the paper.

keywords: FPGA, SIMD architectures, early vision