Hough transform implementation on a reconfigurable highly parallel architecture

Author

Mahmoud, Meribout ; Nakanishi, Mamoru ; Ogura, Takeshi

Author_Institution

Integrated Inf. & Energy Syst. Lab., NTT Human Interface Labs., Kanagawa, Japan

fYear

1997

fDate

20-22 Oct 1997

Firstpage

186

Lastpage

194

Abstract

A highly parallel hardware architecture for Hough Transform (HT)-based parametric curve and end-points extraction is described. It´s based on CAM (Content Addressable Memory) concept. Through using advanced VLSI technology, to classical CAM implementations, three main features are provided: Parallel write, parallel search and single/multi hit response flag. Moreover the high number of CAM words available per chip synchronized with a high frequency clock has the merit to keep low both the hardware amount and the execution time, while featuring the ability to extract efficiently curve parameters and their corresponding end-points. To evaluate its performance, experimental results for segment extraction are presented: By using two CAM chips, segment extraction of an image 256×256 is achieved in real-time

Keywords

Hough transforms; content-addressable storage; feature extraction; image segmentation; parallel architectures; reconfigurable architectures; CAM; Content Addressable Memory; Hough transform; end-points extraction; parallel architecture; parallel search; parallel write; parametric curve; real-time; reconfigurable; segment extraction; single/multi hit response; CADCAM; Computer aided manufacturing; Computer architecture; Data mining; Hardware; Humans; Image segmentation; Parallel architectures; Shape; Very large scale integration;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer Architecture for Machine Perception, 1997. CAMP 97. Proceedings. 1997 Fourth IEEE International Workshop on

Conference_Location

Cambridge, MA

Print_ISBN

0-8186-7987-5

Type

conf

DOI

10.1109/CAMP.1997.631947

Filename

631947