Archived posting to the Leica Users Group, 2000/11/05
[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]At 11:59 PM 11/4/00 -0500, Martin Howard wrote: > >Nevertheless, none of the above has anything to do with Jim's assertion that >autofocus is no good because the programmer who wrote the code for the chip >doesn't know what you're going to take a picture of. Again, it has about as >much relevance as the statement that the guy who paints the shutter speed >numbers on your Leica doesn't know if it's a sunny day or raining when >you're taking pictures at your daughter's graduation. > >Martin Howard | Actually, it does have much relevance. I just finished both autofocus and autoexposure microcode that will be running in a high-end Agilent/RCA camera. Personally knowing the mathematics involved in making a lens move, gathering data and processing it at a high rate of speed, generating histograms of each "focus" capture and correlating it to the previous capture, and using proportional/integral algorithms to speed-up and slow down the lens as it moves to/away from optimum mathematical convergence, I can tell you that the whole process is prone to subject failure. AF can only look at contrast differences. The algorithms implemented will be different for each programmer. Like color balance. In digital cameras we have Asian and non-Asian color balances. We also have Asian and non-Asian focus/exposure algorithms. So, it does indeed make a lot of difference who is implementing the exposure/focus/color balance and it is true that these algorithms are purely mathematical. Which is why you can get AF to hunt endlessly if pointed at the wrong subject. High end cameras such as the D1, F5, EOS1, etc, have many years of experience behind them and do a reasonably good job. you can still make them hunt and still get their exposure algorithms to fail. But it still boils down to that you are the photographer and the camera has no clue what your photographic vision is. It is subject blind and relies on contrast differentials to focus. Exposure gives weights to various matrix configurations (adjustable from nine to 36 in our camera) but still boils the total accumulation of these weighted sectors, to yes... roughly 18% gray. These systems are simply dumb. Photographers give them too much intelligence. Even those using fuzzy logic and AI use math to interpret contrast ratios and gray densities. Humans are much smarter than this. This is why the same lens touts a higher lp/mm resolution if hand focused vs if auto focused. Jim