In 2005, the prestigious Science magazine posted a list of the greatest 125 unsolved scientific questions. Thousands of researchers were disappointed not to find their projects listed. However, in my prejudiced opinion, the fact they didn’t mention vision shows how little the general public heard about it. This is understandable, since computer vision is a young field, about 30 years old. It is so young that nobody had written a thesis on the history of computer vision, since most people who created the field are still alive. No doubt this thesis will be written, when people will realize the social implications of cameras. Only recently we became surrounded by cameras, for example in security applications or mobile phone. Imagine that in your classroom at school there were cameras recording everything to help the teacher. In the US they already started experimenting this in special classes. Sometimes I see good people working on unimportant, uninteresting, esoteric research projects. I believe they would be doing vision research if only they had cameras installed in their classroom.
When People hear about computer vision the first time, they have all kind of associations. Somehow, many associate it with graphics. While both areas have things in common since they deal with images, in some sense vision is the inverse of graphics. Whereas in graphics we have a model of the scene that we would like to render, in vision we observe an image and try to build a model of the world. That’s not to say that problems in computer graphics are easier, since in computer graphics people expect higher visual standards.
People who had Fourier-transform trauma in university often associate vision with image processing. Image processing is only low-level vision. Some processing operations, like image compression, don’t even attempt to understand what is in the image. Other operations, like edge detection, are preparations for higher level analysis. Computer vision is much broader, with mid-level areas like perceptual grouping, analysis of surfaces and motion. And there is also the higher level of recognition and classification.
Other people I met associate vision with robotics. But robotics is just a single application area of vision. There are many others. Unfortunately, many people consider vision mainly as an engineering problem, where the goal is to build systems by composing known building blocks. However, computer vision has also a scientific aspect, where people are searching for principles that may become new building blocks in the future.
The good news is that more people get interested in vision. There are very good books today, and plenty of accessible material on the web. So the next time Science will post a list of unsolved questions, possibly by its 200 anniversary, the vision problem will surely take one of the first places. Unless, of course, if it will be solved by then.
