What's the best estimate for largest detectable object based on pixel size?

[DavidOP]

Let's assume that you take a digital picture of a small object from about 100 feet away. Also you know the exact distance, focal length, so that you can compute pixel size
PS in inches/pixel for the distance corresponding to the object. What is the best estimate for the largest detectable object at this pixel size PS. Is it PS, or 50% of PS,
or less.

 

[Benco]

It mainly depends on the quality of the lens, google 'circle of confusion'. The size (in pixels) isn't straightforward as the size of individual pixels vary from camera to camera, anyway even if you've got a fantastic lens the nature of the bayer array makes going down to idividual pixels pretty irrelevant as far as detail goes.

 

[Garbz]

That depends on the diffraction limit of the system. In theory there's a finite limit for the length of a lens and it's aperture after which a bigger lens won't get you anywhere without a bigger sensor. But in essence your question is missing a few variables.

 

[bratkinson]

While I am not even close to being an optics engineer, as indicated above, there's no 'one' answer.

For starters, the resolution-power of a lens between two separate objects at a distance comes into play. For example, when driving at night on a straight, flat road, at what distance do the headlights of an approaching car go from a single blob of light to two discernable lights? If you are looking through your windshield versus standing alongside the highway, that distance will be different, due to the light-handling capabilities/limitations of that particular windshield. The windshield on a different make/model of car would give different results.

Then, throw in the differences between camera lenses. Say, for example, a Canon 50mm f1.2L vs a Canon 50mm f1.8. The 'resolving power' of those lenses varies as well.

Now consider the camera sensor. As expected, the size of each sensel and the distance between them directly affect what the camera 'sees'.

Now, throw in light-source intensity, such as a search light vs small flash light, light diffraction by wavelength differences within the lens itself (eg, violet light behaves differently than red light as it passes through a piece of glass, etc), even issues of air quality between light source and sensor can affect the outcome.

Bottom line, unless you have an optical laboratory set up, trying to measure the smallest detectable object will be an exercise in futility.

 

[Helen B]

You have phrased the question about object size purely in terms of the pixel size, so I assume that is how you want the answer - ie you wish to find the limit based solely on pixel size so that you could say, for example, "Based on sensel size, at that distance, an object less than x inches across could not be resolved even if all other factors were favourable."

If you do ignore lens resolving power, diffraction, focus etc, there are still reasons for not being able to give a single answer. Do you want to take the effect of the optical low pass filter / anti-aliasing filter into consideration? Suppose that the (ideal) image of the object did not fill a sensel, it could still make a detectable difference to the signal, but the detectability of the difference would depend on the size and contrast difference of the object, and the difference in signal caused by the object's surroundings and by noise (ie the difference in signal caused by the object may not be any greater than the local differences that would occur if the object were not there.

 

[KmH]

Pixels (photodiodes and their associated electronics) generally have some space between them and are on a size scale that is in micrometers (µm - one-millionth of a metre (or one-thousandth of a millimetre, 0.001 mm, or about 0.000039 inches).

On a practical basis, a single pixel cannot contain detail beyond a single luminosity value because all a pixel does is develop a single voltage based on how many photons hit it during an exposure.
Detail can only be rendered by combining a lot of single pixel voltages.

 

[amolitor]

Based on previous posts, I suspect that DavidOP primarily wants to "prove" something about the Zapruder film. If he keeps it up much longer I'll direct him to some debunking material.

 

[amolitor]

If I take a picture of, let's say an open-topped car, from, say, 100 feet away with a camera that is the equivalent of, say, a 640x480 digital imaging system.

Let us suppose that in that open topped car someone were to have an object 0.1mm across. Suppose that object were to emit an intensely bright light directly in the direction of the camera. We would most certainly be able to detect this object in the completed film.

Suppose, on the other hand, than a flat alien creature, several feet across, were clinging to the hood of that open-topped car. Suppose that the color and tonality of that alien creature exactly matched the color and tonality of the hood of the car. We would quite likely NOT be able to detect that creature, despite it being several feet across.