How well can a DSLR retake the same photo?

[Kythoon]

Hi everyone,

I recently started my Master's in engineering that involves DIC (digital image correlation). Basically I can track how a patch of pixels in an image moves between consecutive images and relate that to real world movement.

As the program/software gets more precise (sub-pixel level) I started wondering about how well the camera (Canon T1i) can retake the same image. Does anyone have any ideas about this? Does RAW vs jpeg have any effect? Wondering if the applied sharpness might vary between different images.

I'm currently testing the mirror lock-up to see what effect it has.

Thanks for any ideas

Yes, RAW vs. JPG makes a world of difference. Especially at the minute levels.

RAW vs JPEG isn't really a comparison. RAW is the actual output from the sensor. JPEG is taking the RAW file, analyzing it, compressing it, and spitting out a smaller file. When you get a JPEG out of the camera, the camera does the compressing with it's built in software. It's the same as if you were to create a JPEG in Lightroom out of a RAW, except that the user has some control over how the JPEG is formed.

To answer your question, yes, a DSLR can take the same photo twice. However, that would also assume identical conditions between both photos, and the same light is hitting the sensor in the exact same spot. Variations of light from sources can be small, but can affect the outcome of the photo slightly.

There probably is a built in tolerance of the sensor as well. Each pixel is probably accurate to a certain degree, and could possibly vary from shot to shot. Temperature seems to be the biggest factor in semiconductors, but other factors (such as number of uses) can impact the results as well.

[sk66]

I think if you can SOLIDLY lock down the camera so no movement is possible then it could take the exact same image with the only "movement" being that of the subject. I don't think that mirror lockup would be necessary. I would use a prime lens and a fixed aperture...the only remaining factor would be how stray "unfocused" light bounces thru the lens and is recorded. Avoid backlit subjects as CA can be recorded erroneously.

There is the fact that each site on a CCD sensor can only record one color, so if red hits a green site nothing is recorded.a program then interpolates what color should be displayed for the sites where nothing was recorded. So if you are determining movement by the shift of color in an image (I don't know what eles you could be measuring) your accuracy would be limited to within 3 sites (for a Bayer RGB sensor). You could get a camera with a Foveon Sensor.

[Kythoon]

Thanks everyone!

Yes I knew about the RAW/JPEG difference, just wondered if the sharpening, for example, would change between "identical" photos.

The lens is a Canon 180mm macro and I'm shooting in manual mode and manual focus. Taking the images in a classroom with the windows and door closed, using just overhead and natural light.

That is an interesting point about the sensor only recording a certain colour. In the software processing, the images are converted to gray scale but nevertheless this would still be an issue.

In terms of mounting, I have tried both a tripod and just balancing the camera on the lens tripod mount on a desk. With the tripod I have been noticing a drift downward over time (gravity) even though I have tightened it as much as possible). Otherwise the general trend is the camera shifting to the right; as well as the expected random noise.

[ravncat]

The shot noise is random - there should be a different shot noise in each shot, (this is why stacking multiple shots can "lower it's appearance". Lower ISO can make it all but disappear, but I suspect very minute tolerances should detect the differing patterns , especially at the 12 or 14 bit depth - even at the lower ISO settings. I think getting two exactly duplicate exposures / compositions is possible - but the noise (read noise / heat noise) should produce things that would be "forensically" different.

[NicoleScraps]

I don't know if this will help but my camera has a RAW+ Jpg option. It will take take the same photo and save it twice, once as a jpg and once in RAW. Maybe yours does too? I am not even sure that I thoroughly understand this thread in the first place, so just ignore me if my response makes no sense.

[Kythoon]

I do try and lower the ISO to help reduce the digital noise; but knowing that it's essentially random each time, I'd like to see if the other variables are controllable.

I'm guessing I'd need a different camera to shoot at a higher bit depth - currently only in 256 colours. This true?

Besides the mirror flip, what else moves inside the camera during an exposure? Wondering what else could minutely shift the image (talking 0.01 pixel and smaller range here).

[ravncat]

"Besides the mirror flip, what else moves inside the camera during an exposure?"

The shutter itself, of course. It`s usually a three step move, curtain one comes down - followed by curtain two, and then the reset.

During the mirror flip - a dslr often also has a secondary mirror - this is in step with the main mirror and has to do with the phase autofocus system

Otherwise - I don`t think you`ll have any other moving parts - unless of course you`re using a sensor with image stabilization - then the sensor moves too .

Lenses can have moving elements for Image stabilization - They could affect how light falls onto the sensor. Likely you wont have moving elements or aperture blades during a shot. (lest you are using a camera with a leaf shutter (mostly medium and large format)


--

Jpg is 8bit. Most cameras shoot in 12bit for RAW capture. Some higher end ones can shoot 14 bit.

[Kythoon]

haha right, the shutter does but was wondering what else. I remembered reading about turning IS off when on a tripod so I have done that already, but good point

I'll let you know what I find in the coming days after I try the RAW capture and the other ideas.

Thanks