<title>Using Microprocessors For Image Processing</title>

Abstract

AbstractThe low cost and simplicity of microprocessors will enable the image -processing fieldto expand widely. Data acquisition, conversion, storage, and display are becoming similar-ly inexpensive. Described is a low -cost image -processing system that illustrates the useof microprocessors. Also, a set of high -level image -processing primitive functions thatare machine independent are proposed. These primitive functions may be used to write high -level languages in less space for shorter execution time. Their machine -independencyshould encourage widespread exchange of programs among members of the image -processingcommunity.IntroductionA decade ago, image -processing research was confined to specialized laboratories havingelaborate scanners and generally poor image- reproduction devices. The major stumblingblock was the existing digital computer. Efficient image processing requires extensivememory, mass storage, and a relatively high -level language, none of which were readilyavailable in minicomputers. When left to large -scale computers, a different set of prob-lems arose. Unless the researcher were lucky, operations were restricted to batch or time-sharing modes due to the great expense of on -line, real -time computation. This meantblindly processing several swatches of imagery before playback.All this has changed dramatically. Today, minicomputers are available as dedicated,on -line equipment, with the hardware and the high -level languages necessary to do the job.Combined with improvements in image acquisition and recording /display devices, minicom-puters have enabled the image- processing community to expand rapidly. Changing technologyis now permitting yet another generation of image- processing hardware; microprocessors,combined with solid -state detector arrays for image acquisition, will make commercial andeven hobbyist image- processing a reality within the next few years. To the hobbyist, com-puter picture -processing (especially color) promises all of the excitement that amateurradio or photography did fifty years ago. To industry, computerized image -processingoffers a wide variety of cost- effective applications ranging from quality control of pieceparts to robot parts- assembly based on pictorial pattern recognition.System ConfigurationIf constructed today, a minimum- hardware, microprocessor -based image -processing systemmight be configured as shown in Figure 1. A TV camera, either array or conventional, wouldprovide a pixel representation of the original scene. Array cameras have the advantageof being blocked as individual pixels but are now somewhat more expensive. However, theirpotential for mass production and availability for color acquisition will soon make themattractive alternatives to conventional TV cameras.Bulk storage on flexible (floppy) disks is presently lowest cost storage. Flexibledisks will soon be replaced by faster magnetic- bubble memory. Main memory and refreshmemory have declined in price rapidly enough to make the cost of commercial image-proces-sing decline still further. The remaining electronic blocks shown (such as the A/D con-verter, Direct Memory Access (DMA) controller, CRT controller, floppy -disk controller,and D/A converter) can be purchased as LSI entities of one to a few chips each.In the example shown, a reasonable image size might be a 128x128 8 -bit image. For thissize and encoding, an A/D converter sample rate of 2 MHz is required in order to scan atstandard video rates. Likewise, the DMA controller, the refresh memory, and the CRT con-troller all require a 2 MHz throughput rate. This video rate is well within present state -