用FLEX10K10实现小型卷积处理器

Implement Minimum Convolution System with FPGA FLEX10K10

当对图像进行滤波、增强、边缘探测等处理时需要用到卷积运算 ,专用蕊片可以实现实时卷积运算 ,但价格昂贵 ,普通场合不易使用 ,随着大规模可编程逻辑器件 (FPGA)的出现 ,包括卷积在内的许多算法硬件实现变为可能。该文基于ALTERA公司的MAXPLUS开发软件 ,选用FLEX10K10器件 ,设计了一种能够对视频信号进行实时处理的卷积器 ,卷积窗口 3× 3,数据宽度 4bit× 8bit,延时2 0 0ns。

In imaging processing such as filtering, enhancing and edge detecting, 2D convolution is often used. Convolution algorithm can be implemented in real time with some special chip.But its price is so expensive that it can not be used on general occasion.Along with the appearance of FPGA,many algorithm including convrolution can be implemented with ease.Based on the MAXPLUS system of Altera Corporation, a convolution implementation scheme of video in real time is proposed in this paper. Its convolution window is 3X3; data width is 4bit×8bit Some advice has also been given here to be considered in design.