Improved Sequential & Parallel Designs and Implementations of the Eight Direction Prewitt Edge Detection

Mohammed B. Mohammed
University of Colorado
University of Colorado, 2013





   school={University of Colorado}


Download Download (PDF)   View View   Source Source   



The exponential growth of the world’s technological industry has an important impact on our lives; we are witnessing an expansion in computer power combined with a noticeable development of digital camera capabilities. To keep up with the requirements of the digitalized world, the focus has been set on the computer vision field. One of the most popular computer vision applications is recognition. The word recognition can imply different computer vision areas such as object tracking, face and pattern recognition, human computer interaction, traffic monitoring, vehicle navigation, etc. Edge detection algorithms are widely used within the computer vision and the image processing field. Edge detection algorithms are at the center of the recognition process in computer vision and image processing. This work presents design and implementation of efficient sequential and parallel edge detection algorithms capable of producing high quality results and performing at high speed [1]. The parallel version, derived from our efficient sequential algorithm, is designed for the new shared memory MIMD multicore platforms. The edge detection algorithm presented here is designed to effectively work on images impacted with different noise percentages. This has been achieved through augmenting our edge detection algorithm with an improved median filter capable of suppressing impulse noise and other noises more effectively than the original standard Median filter. A global thresholding method augments our design to dynamically find a suitable thresholding value. In order to measure the quality and execution time, we test images with different sizes along with the original Prewitt and Canny edge detection algorithms already implemented in Matlab, to show the possibility of using our design within different applications. This work will demonstrate the ability to process relatively small and medium images in real-time as well as effectively processing extremely large images, useful for biomedical image processing, rapidly.
No votes yet.
Please wait...

* * *

* * *

* * *

HGPU group © 2010-2022 hgpu.org

All rights belong to the respective authors

Contact us: