Solve complex vision problems by seeing things as they are

"We see things as we are, not as they are" – Leo Rasten

Human beings have a good ability to focus on the task at hand. This is a learnt behavior that is developed since childhood. As the tasks become more complex, we learn to focus to master the new ability and to finish a new task. At the same time, we humans have a tendency to be self absorbed when we focus on our work. We would overlook overtly simple patterns that might accompany the new complex task.

I have previously noted that dangers of biased approaches to vision. That is, we have a tendency to naturally design vision systems based on our own visual system. This inclination is natural and it takes a new ability to find new approaches. For example, our imaging systems have been based on arrays of red, green, and blue pixels, much like the retina in our eyes. Are these the best approaches for vision? We have designed distinct capture and processing system (imager array and processor) much like our eyes and the brain. Is a distributed processing system, one that co-locates the processor with pixel, a better approach? These are all interesting research questions for us to consider.

We might want to take time to step back and reconsider the problem at hand. Over time, we have been so focused on the research task at hand that we could have biased our ability to consider other approaches. It takes a new level of ability to be able to see the different approaches for computer vision problems.