A NEURAL-NETWORK ACCELERATOR FOR IMAGE-ANALYSIS

被引：2

作者：

COSATTO, E

GRAF, HP

机构：

来源：

IEEE MICRO | 1995年 / 15卷 / 03期

关键词：

D O I：

10.1109/40.387680

中图分类号：

TP3 [计算技术、计算机技术];

学科分类号：

0812 ;

摘要：

At the heart of this image analysis system are two NET32K analog neural network chips, computing over 100 billion multiply-accumulates per second. The system simultaneously scans sixty-four 16x16-pixel templates over bi-level images, producing feature maps that mark matches between the image and the templates. When we code simple, generic shapes into the templates, the feature maps allow us to make quick, robust analyses of complex, noisy images.

引用

页码：32 / 38

页数：7

共 11 条

[1]

BROWN CM, 1982, COMPUTER VISION, P65

[2]

Dubes RC, 1988, ALGORITHMS CLUSTERIN

[3]

GRAF HP, 1990, ISSCC, P144

[4]

GRAF HP, 1994, NEURAL INFORMATION P, V6, P785

[5]

LYLE JD, 1992, SBUS INFORMATION APP

[6]

PERSOON E, 1988, IEEE T PATTERN ANAL, V10, P338

[7]

PRATT WK, 1991, DIGITAL IMAGE PROCES, P651

[8] THE ARCHITECTURES AND DESIGN OF A 20-MHZ REAL-TIME DSP CHIP SET [J].

RUETZ, PA .

IEEE JOURNAL OF SOLID-STATE CIRCUITS, 1989, 24 (02) :338-348

[9]

Srihari S. N., 1989, From Pixels to Features. Proceedings of a Workshop, P261

[10]

1994, ELECTRONIC DESIG MAY, P136

← 1 2 →