FPGAs: On a mission to help the color-blind

Over the last few years, I've become very interested in biological and robotic vision systems. For example, at some stage along the evolutionary path—let's say 800 million years ago—some multicell organisms managed to develop photoreceptors that gave them the ability to detect and respond to some form of light.

It may be that these first photoreceptors were cone cells that were primarily sensitive to light in the UV portion of the spectrum. (Cone cells are so named because of their shape under a microscope.) Alternatively, the first photoreceptor may have had its peak sensitivity way up in the red region at the opposite end of the spectrum. Both scenarios are consistent with existing data. In any case, during the next several hundred million years, the creatures that were to evolve into vertebrates, dinosaurs, mammals, primates and—ultimately—humans developed four different types of color photoreceptors/cone pigments. Thus, these creatures would be known as tetrachromats.

Sad to say, however, our ancestors lost first one and then two of these pigments sometime between 310 million and 125 million years ago. We don't know exactly when or why, although one possibility is that these creatures became nocturnal. This explains why most of today's mammals are dichromats with only two types of color photoreceptors.

Stick with me, we're almost home. Sometime between 45 million and 30 million years ago, the primates that were to evolve into humans "split" one of their color photoreceptors into two types. Thus, typical humans have three types of color photoreceptors and are known as trichromats. This gives us a much richer visual experience than our dichromat cousins (such as dogs and mice) who tend to perceive images only in terms of blue and yellow.

Daltonize-type algorithm
So, why am I waffling on about all of this? Well, it may be that anywhere between one in 12 to one in 20 people are color-blind in one form or another. Approximately 1-2 percent of all men on the planet are completely missing one of their three color cones. Another 7 percent have all three cones, but one or more are abnormal in some way such that they aren't able to receive the same amount of color information as the rest of us.

The Vischeck software at www.vischeck.com allows you to run your own images to see how they appear to people with different types of color deficiencies. Of particular interest is that the folks at Vischeck also have a program called Daltonize. This program is really clever. If you tell it what form of color deficiency you have and provide it with an image, it will analyze the image and replace the colors you can't see with other colors you can. And it accentuates details that would otherwise be hidden from you while it does this.

So, my idea is that it would be really useful if someone looking at a TV or computer monitor could tell the display about any color vision problems he or she has. Then the display could use Daltonize-type algorithms to correct for this on the fly. In the case of TV, of course, you would probably leave it on the "normal" setting if multiple viewers were enjoying a show.

Meanwhile, in the case of computer displays used by individuals, I think this capability would be a no-brainer. If one display manufacturer had this technology, a lot of users would be interested. This manufacturer would sell more units, driving the other manufacturers to follow suit. In fact, this could be advantageous for any form of graphics display, such as GPS in cars and trucks.

What would be a good way to implement these algorithms? Well, you would need massive parallelism and humongous amounts of processing power, which scream FPGAs to me. Of course, this may never come to pass. On the other hand, it may be that the "egg heads" working in the top-secret underground bunkers at the FPGA companies are working feverishly on this as we speak.

- Clive Maxfield
Editor, Programmable Logic DesignLine