Genetic Algorithm Based Design And Implementation of Multiplierless Two-Dimensional Filters

Union College CPe-198/199 Capstone Design Project

Christopher D. Roblee                                           Douglas J. Lockett

 

Documentation

Overview

Motivation

In today’s fast-paced society, effective real-time processing technology is essential. The prevalence of digital multimedia has driven the development of such technologies. Modern applications of signal and image processing demand substantial levels of computation, many of which are too complex for practical software implementation. Software is slow, and limited for real-time applications.

DSP/Image Processing

Why do we choose to process images? Images are everywhere The concept of sight/vision is both complex and fascinating Can be perceived as 2D discrete signals. Many of the same traditional rules and applications of DSP can be applied to Images/Image Processing. Our Focus: Edge Detection One of the most common forms of image data extraction. The information from detected edges in images can be used in a variety of different applications.

How are Image Edges Detected?

One way is to apply a 2D-high pass filter to the image. Sharp discontinuities in the brightness of the image appear as high frequency components in the image. These sharp discontinuities are the “edges” in the image. Will most likely be realized as an IIR filter IIRs (Infinite Impulse Response) filters generally require less computation than FIR filters and need a smaller order. They also allow a sharper frequency response.

Example of Edge Detection Technology

IIR Filtering

Filter Characteristics Must satisfy characteristics of high pass edge detection filter. All filter coefficients must be base-2 to satisfy multiplierless criteria. Filter must be stable!

Filter Design

Once desired filter type has been chosen, an algorithm must be determined to design the filter based on the characteristics. One possible method is to develop a genetic algorithm to determine base-2 coefficients such that the filter remains stable. Filter design will be done in MATLAB

Hardware Implementation

VLSI Chip will be designed and fabricated to perform the complex 2D Image filtering operations. The chip will contain memory to meet the recursive requirements of IIR filtering.

Benefits of VLSI Chips

The VLSI (Very Large Scale Integration) hardware design methodology is used throughout industry to implement a wide variety of complex computing applications. Fast Reliable Minimizes die size and system power consumption.

Constraints

Testing, evaluation and re-design -Time consuming, exhaustive. -Top-down and bottom-up approaches necessary. Budget -Use MOSIS IC Prototype Fabrication Service Overall scope of applications

Students:
 Chris Roblee
 Doug Lockett
Advisor:
Professor Michael Rudko