A Classified and Comparative Study of 2-D Convolvers
Paper ID : 1101-MVIP2020
Mahdi Kalbasi *1, Hooman Nikmehr2
1Department of Electrical and Computer Engineering
2Department of computer engineering, University of Isfahan
Two-dimensional convolution is a common operation in a wide range of signal and image processing applications. Although most commonly used in real-time applications, the convolution is a compute-intensive operation. Moreover, it is a memory-intensive operation and memory access is an important challenge of a 2-D convolver. This paper introduces a new classification of the most important and commonly used convolution designs in four different classes and then the advantages and disadvantages of them within these classes are discussed. The results of real-world implementation of these designs using Verilog on a Virtex-7 FPGA are represented and compared in terms of the critical path, memory bandwidth, and resource utilization.
2-D convolution, Compute-intensive applications, Memory-intensive applications, Resource utilization
Status : Paper Accepted (Oral Presentation)