Image Processing and Computer Vision

Ever since the invention of the iconoscope image processing has been tremendously important. The early work here was very much in the spirit of the Fourier analysis of 1-D signals (sound signals), filtering, digitization, etc. The way that TV sought to broadcast images was to scan the image and use the resulting time domain signal as the modulating signal. Even for conventional TV this is a high bandwidth signal (4.25 MHz), it is of course much higher for High Definition TV (HDTV). Thus the search was on for image compression and coding techniques. This is a topic of much research to this day. In parallel with this activity began an activity in pattern recognition, where the goal was to automatically read characters and to recognize patterns for surveillance, inspection, etc. While recognition of type written characters is easy now, handwritten character recognition is not yet easy.

Computer vision came to the fore much more recently in an attempt to look at images in some fundamentally new ways:

1. Semantic content of the images: one wanted to take into account the meaning of the image before processing it.
2. Oculo-Motor Control incorporating the vision sensor in a feedback loop for a robot or a car or an airplane.
3. Biomimetic Information Processing trying to mimic the working of the mammalian vision system