March 23, 2009 – A brief history of digital imaging from one man’s perspective.
By The Cerebral Aesthetic Vagabond
The other day, in response to my father’s enthusiasm, I looked up the Canon PowerShot G10 digital camera, and after my salivating over its 14.7 million pixels subsided I reflected on how much digital imaging has evolved in my lifetime.
About twenty-five years ago – good grief, a quarter of a century just doesn’t sound as long as it used to! – electronic imaging largely consisted of video images recorded onto magnetic tape. VCRs and video cameras were pretty well established by then and quite affordable. However, digital imaging hardly existed, except perhaps in research laboratories or government facilities. (Laser disc video technology existed then, but it was a read-only technology and quite expensive.) In the mid-1980s I worked for a small electronic engineering firm which had a lot of experience designing plug-in boards for Apple II computers, and later for IBM-compatible PCs. One day my boss became intrigued by the notion of digital imaging – he was always a “seat of the pants” kind of guy – and we immediately embarked on a development project bring low cost image compression to the PC platform. Mind you, we were only interested in developing image compression technology, not in finding ways to actually utilize image compression technology. That was to come later. In the meantime, we all harbored a sort of blind faith that we would find some use for this technology, especially since our boss was the one taking the financial risk!
At the time, an image capture board that could be plugged into a PC cost about $4,000. It accepted a video signal from a video camera (another couple of thousand dollars), captured a single video frame and stored it on the computer’s hard disk drive. This image was not compressed, but stored as raw RGB pixel data. As memory serves, these uncompressed images offered only about 32,000 pixel colors (5-bits each for red, green and blue) and required about half a megabyte of disk space. We scoff at such minuscule storage requirements today, but back then the typical computer had a 20 megabyte hard disk drive! At that time, my computer at work – I was the software guru, after all – had two whopping 30 megabyte hard disk drives, which if used solely for storing these captured digital images would have permitted me to store a grand total of about 120 images. Contrast that with the Canon camera mentioned above, whose web site page claims that as many as 1,200 images of the highest quality can be be stored right inside the camera, each with on the order of fifty times the number of pixels of those back in the 1980s! Needless to say, hard disk drives today can store not mere hundreds, but hundreds of thousands of such images.
The massive storage requirements of such images back then was only one limitation we sought to ameliorate. The Internet, while it existed back then to serve the government and its university researchers, was not accessible to much of the public, nor did “high speed” computer communications exist. “High speed” back then meant 14,400 baud analog modems over conventional telephone lines, a speed that would require upwards of five minutes to transmit just one of those uncompressed digital images! Our lofty goal of compressing those digital images aimed at increasing the storage capacity of our hard disk drives and the speed of image transmission by an order of magnitude.
But there were some hurdles to overcome first, among them the fact that most PCs had 16-bit microprocessors (only 8-bit data buses, though) running at 4.77 MHz and 640 KB of RAM. Compare that to today’s computers which typically boast 32-bit data buses, microprocessors 500 times as fast and memories 5,000 times as capacious. Because of the paltry computing resources at our disposal we chose to develop a co-processor board containing a digital signal processing (DSP) chip that would plug into the PC and perform the “heavy lifting,” the fast Fourier transforms that formed the heart of the image compression and decompression process. Software running in the PC would slice up the uncompressed image into tiles, send the tiles to the co-processor board to be FFTed, and then receive them back for further processing. To achieve maximum compression and decompression speed, it was crucial to run the two microprocessors – the native CPU and the DSP – in parallel, so we developed a fairly complex and elaborately choreographed software design that maximized the parallelization of the operations, without even the benefit of a modern multithreaded operating system. Our software was all done in the C-language running on DOS on the PC side, and in assembly language on the DSP side.
We succeeded in our goal, however, and achieved good quality compressed images one tenth the uncompressed size. Then we set about finding markets for our “low cost” $1,000+ co-processor board and its accompanying software tools. Bear in mind that $1,000 was about half the cost of the entire computer back then, and those were 1980s dollars. In today’s dollars, everything would be two to three times as much. Surprisingly, we found quite a bit of demand for our product, at least for the first few years. Once 32-bit microprocessors started to become commonplace in desktop computers beginning in the early 1990s, it was clear that our little niche was doomed, as the desktop computer’s microprocessor was capable of performing the entire image compression process itself, Fourier transforms and all, without need for an expensive co-processor board.
Today digital imaging is ubiquitous and powerful microprocessors with enough computing power to perform image compression are found in everything. Digital imaging can be found on computer screens; whizzing across wired and wireless communication channels; on televisions; in digital cameras that have all but replaced film cameras; in Dick Tracy-style personal communicators, otherwise known as mobile phones; atop traffic light poles at intersections; on the dashboards of cars equipped with rear cameras to assist in backing up; and in a hundred other places, not all of them benign. Almost all digital imaging today utilizes image compression technology similar to what we pioneered on lowly PCs back in the 1980s. As with so many other technologies, today we take for granted zapping a multi-megabyte image from our mobile phone to someone, oblivious to the groundbreaking effort that went into it.
Like all technological developments, the digital imaging revolution has been used for good and bad. It’s amazing, for instance, to be able to send a digital image of someone’s x-ray to a doctor far away for immediate analysis. It may be unappreciated by domestic health care workers, however, if that doctor happens to be in another country! It’s wonderful to be able to snap a photo of the kids and send it to their grandparents thousands of miles away, at virtually no cost. It’s not so nice to have a robot decide you’re speeding and snap a photo of your license plate and automatically send you a bill, until the receipt of which you continue to live in blissful ignorance of your transgression. It’s nice for us people to be able to see satellite photos of any interesting place on earth. It’s not so nice to realize that governments can do the same. It’s nice to be able to whip out our mobile phone, film the police misbehaving and promptly post the resultant video on the Internet. It’s not so nice for governments to criminalize such citizen activism while they themselves digitally film citizens engaged in peaceful protest.
It’s just remarkable how rapidly digital imaging technology has evolved. In some ways it has perhaps cheapened the “art” of photography by transforming photographs into abundant commodities. On the other hand, thanks to digital imaging we can do things today that were undreamed of even a couple of decades ago.
Oddly enough, I’m something of a Luddite when it comes to digital imaging. I don’t have a mobile phone with a camera in it, nor do I want one, and my digital camera is seven years old and captures a mere six million pixels. The irony of my salivating over a 14.7 million pixel camera is that I seldom make use of images containing more than half a million pixels. Nevertheless, if I come into some money, I might just splurge and buy one of those cameras.