An assumption of knowledge
The cognoscenti are blowing smoke while not delivering digestible information about machine vision.
Being a journalist is both a rewarding and frustrating experience. Luckily the many benefits of this fine and noble profession often outweigh the frustrations. Being able to travel to exotic locations such as the Donald E. Stephens Convention Center in Rosemont, IL, USA, and the equally picturesque National Exhibition Centre in Birmingham, England, is just one of the many perks of the job. More important, the job provides a chance to meet some very clever people and interrogate (that is, interview) them about new technologies, software algorithms, and future products. In doing so, one can learn a great deal about the subject in a very short time.
Even after traveling long distances to faraway places, things do not always go as planned. At one recent trade show, I met representatives from a very large Japanese camera manufacturer. Always keen to obtain a story, I asked about the design of the product, what possible future generations would look like, and whether any customers in the United States were willing to discuss their application. While no one knew the answers to any of these questions, the company representatives promised that someone would e-mail me with the information. Unfortunately, no-email was ever delivered.
Sometimes, however, the people whodo know the answers to specific questions prove equally unhelpful as those who know nothing at all. This came to light last month while I was researching an article on image-processing techniques based on partial differential equations (PDEs). Of the many web sites, technical papers, and application notes posted on the Internet, several assumed prior knowledge of the difference between PDE-based techniques and more conventional kernel-based filtering. Many authors described mathematically how such PDEs were formulated without any explanation of underlying assumptions about the process. Worse, they provided no examples of how such algorithms could be applied to images or any benefits over conventional approaches.
Of course, in writing these papers and tutorials, an author must assume a certain prior knowledge of mathematics, physics, and computer programming. But at what level should this be? Since I have a degree in device physics (with third-class honors, no less), I would like to think that these texts are written so that any person with a degree in mathematics, physics, or computer science would understand.
Unfortunately, many of these authors seem to believe that they belong to an elite cognoscenti club in which they can sit around in leather armchairs, smoke cigars, and talk at the most abstract level to convince themselves how clever they are. But I pity the engineer who is new to the field of image processing and machine vision-or, in fact, those already involved in the industry.
Better tools needed
Indeed, very few good textbooks exist that present the subject of image-processing and machine-vision systems in a way that encompasses both the mathematical concepts and how to apply them and provide examples of how algorithms can be applied. With the advent of dynamic Web-based communication, there is no excuse for the fact that those involved in education are not providing more effective education tools-tools that address these issues with sound, slides, embedded video, and, more important, understandable content.
Here atVision Systems Design, we like to think that we are playing our part. Our series of Webcasts hosted by Valerie Bolhouse, entitled “Fundamentals of Machine Vision,” has won great praise from the more than 600 viewers that watch each presentation.
For those of you who have yet to view these Webcasts, I urge you to take a look at how Bolhouse’s presentations bring textbook knowledge alive with interesting graphics. It’s a lesson some of those sitting and smoking in the cognoscenti club could well learn.
Andy Wilson
editor
[email protected]