The Real-time Image Filtering Methods Realized By FPGA

Image filtering is the important component of digital processing, this paper design and achieve a complete digital image processing flow based on FPGA,that include image data real-time acquisition、procesing and display.Key realize the fast median filter improved Gaussian filter、image edge extraction image enhancement and image segmentation based on histogram.This paper main introduce the design and implementation for image collection、 cache and display,then detailed introduce various image processing algorithm.The structure of this paper as follows:The first chapter give overview of subject background and research significance.The second chapter introduce the development language、tool and the main process of FPGA,that is base of the follow-up work.The third chapter presents the specific design of image data acquisition、cache and real-time display based on FPGA.The acquisition part is collect image by CCD.The cache part is put1frame image cache to external SDRAM chip.The display part is VGA displayer. A frame image cached by SDRAM play a coordinate role between continuous frame acquisition and display.The fourth chapter realize some low pass filter process for image based on FPGA,that include fast median filter and improved gaussian filter..Rapid median filter make full use of the line technology of FPGA and can find the median in6clock.The method increase the speed of operation and it doesn’t affect real-time display.According to the problem of traditional gaussian filter not only filtering the noise but also lost image details,the improved gaussian filter not only consider the distance between pixel but also consider the difference between pixels when conduct weighted average operation.The fifth chapter realize the real time image edge extraction separately by Sobel、 Prewitt and Laplace operator,analysis and contrast their results.The sixth chapter realize and design the image enhancement and the image segmentation based on histogram.In image enhancement,first process histogram statistic for the current frame,and then process histogram equalization operation in blank time between two adjacent frames,finally,deposit the results to two port RAM for mapping display by next frame.Because the great similarity between two adjacent frames,that doesn’t affect visual result.In image segmentation,first get the histogram of the current frame,then regard the pixels which distribute in5%-95%as background information and regard the others as goal of segmentation extraction.