
I understand the procedure and have even done such calibrations! I guess I should have asked explicitly: how would that be worked into the routine "operation" of a camera? For example, assuming the ...

Looks about right. You need both heading and elevation angle. Magnetic compasses in phones do not seem to consistently achieve anywhere near 3 deg accuracy, especially not without recalibration ...

We haven't progressed beyond the original point. What system can determine the camera's orientation with an accuracy beyond that of the cost and sizelimited systems in a camera?

If it isn't accurate, then it won't help. See above.

I find it very unlikely that your $300 GPS sports watch has a GPS and a magnetic compass with the kind of response time and accuracy you'd need to significantly improve an IS system for the ...

I don't think its quite that cheap and easy to get an accurate magnetic compass, plus they are affected by nearby magnetic objects.


You mean you don't understand how rotation could affect a gyro sensor or just that the magnitude seems off?

No, I do not, because I think it is possible to construct a process that produces a version of an image that contains less (e.g., Shannon) information but appears "sharper" than one containing more.

It was your theorem, you need to provide the definition. There are so many fuzzy terms here. Sharper? Appropriate? If I were allowed to enhance my prints with a felt tip marker, of course I could ...

I'm not even sure what it means to be "captured with less spatial resolution information." However, by most normal definitions of "information," the theorem would seem to be false.

Sure, you could say that, sort of. But strictly speaking, the Nyquist frequency does not set an upper limit, so there are still ways to recover higher spatial frequencies under certain conditions. ...

They are related, but different concepts. Sensitivity typically gives you a threshold for input that gives a specified output, at a fixed SNR. Clearly, if the SNR is allowed to be worse, the ...

Sure, this could be one sensible definition for SNR, but how would this number determine the image data's information content?

This is a really bizarre response. I am saying that if the original "signal" is inherently quantized, then whatever the internal workings of the receiver, at the end of the day we ought to be most ...

What I am asking is where the additional information comes from? What you seem to be saying is that the optimization could use additional information that a) is not part of 'known information,' but ...

" Pick" was your own word. I did say "handwavy" because I thought your reasoning skipped over important steps needed to tie everything together. I apologize if you feel this was an unfair ...

Wait, if the entropy was maximized based on the known information, then how can the result of maximization contain more information than that?

No thanks, as I have never claimed there were more than 2 distinct output values. What I have said is that that those 2 distinct values can represent different measured signals. Well, first of all, ...

2^(bit_depth) is the limit set by the quantization noise floor for the output of an ADC when the input is a certain kind of continuous signal. If the signal cannot be that kind of signal, the ...
Activity older than 12 months is not displayed.

olliess has not added any gear yet.
Total messages 
1317 
Threads started 
2 
Last post 
4 weeks ago 
Photos uploaded 
3 
Last upload 
3 months ago 
