Resampling Images for Comparison

[ameerat42]

If this topic has been raised before, apologies for not finding and adding to it.
If raised before AND by me, even more apologies and a few :os

So it's this - and I will try to state it without any implicit bias...

You want to compare the image quality between two different cameras and their image sizes
are different, say, one is 18MP and the other is 24MP. Do you think that resampling the image,
either up or down to make them the same size, is a valid step in the comparison?

Obviously, you might want to start with as many variables set equal as you can, like the same scene,
same lighting, lens, and if possible, processing.

Also, mention anything else that you think would make a valid comparison.

For people who are newish to resampling, it is the process of using software to up- or down-scale
an image to make the pixel size the same as that of another image. Another description (loosely used)
is "interpolation", but to me that's just one method of resampling.

Ta.
Am.

[MarkChap]

My question is -
Why would you resample ??

You cannot make one camera have more MP or the other have less

You have to look at the images produced by each camera, as they are and make a choice from there

[ameerat42]

Thanks, Mark. I will acknowledge replies with an automated thanks so as to not bias views.

[Mark L]

Going up doesn't help much I'd imagine. If just for displaying on the interweb it might not make a lot of difference, if you're pushing extremes it might.
I don't really have a clue but thought I'd say something to get an automated thanks. I've never had one of them before (I think?);)

[arthurking83]

I've read people's opinions that upscaling in Ps can yield a significant increase in 'image size' .. sometimes of up to 25%.

That is, back in the day when the Nikon D3 only had 12Mp, many pro users were saying that it's images offered unnoticeable quality differences of up to 25% upscaling.
Don't really know about other cameras, but 25% over 12Mp netts you a 15Mp equivalent image.
Could make a difference when printing large-ish.


I think resampling is a valid step and use it all the time to assess what I think a specific camera's sensor is capable of.

That is, I'm thinking of replacing the D300 .. nothing wrong with it, just want to do something else with it.
So I'm always looking at what the gear that I'm interested in is capable of.
eg. I basically try to compare what camera A can do relative to what a D800(my current regular camera) is capable of.

They all eventually have a final output product .. eg. whether that's a print or a digital rendering of some type.
If your purpose is to do the same thing with the two different cameras, then resampling needs to be factored into the equation .. otherwise the comparison isn't really valid.

If your purpose for the different cameras is itself different, then resampling probably isn't something to worry about.

eg. if you're a landscape photographer and your final endpoint is that you always print say a 30" sized print of the final image.
Then if your consideration is to compare the difference between an 18Mp and 24Mp camera, then resampling is important. In the above is a fixed endpoint, and your concern is about whether you're going to see any benefit one way or another between the two different bits of equipment.

BUT! .. if on the other hand your end point for the two different cameras is in itself different .. that is say you want the 24Mp camera for those same landscapes, but you want the 18Mp camera just for macro, or abstracts, or sports .. or whatever .. then one of your other variables has changed, you amy not also want to print those images at 30" .. you may want to print them at say 12 or 13" or whatever because they're easier to sell .. or you may only want to sell online at say 1000px sizes in jpg only format .. or whatever.

The important point is to identify what your endpoint target.
If that purpose changes, then the comparison is needs to be altered to suit that purpose .. etc.


So I think the answer to the question posed is yes, but with a few caveats(as above).
resampling can be important if a specific requirement is needed to be accounted for.

OK, so how I use that above scenario is:

lets say I'm interested in the D500(20Mp) and my current camera is a D800(36Mp) ... and I want the D500 for (lets say specifically) birding/sports/low light faced paced stuff.
I've assessed the cameras ability to shoot rapidly and focus quickly and accurately .. and now discount that in the argument
Now I need to assess if the camera will give me any image quality advantage.
I like using DPRs studio scene tool, as it allows you to roughly estimate if there is any advantage between one sensor and another.
We don't know if the results are perfect, but we can assume that they are ideal.
I can't think of any other site that does this similar comparison feature.
I look at the (say) ISO25K images from each camera at 100% magnified view and this only serves to illustrate to me that if I used the D800 in cropped mode(or cropped heavily myself to a Dx size) .. then the advantage is with the newer D500 if noise quality is the primary concern.
You'd kind of expect this too.
But If I also wanted to shoot landscapes, then I view the entire images as a whole .. that is no pixel peeping. I shoot with an equivalent lens(eg. 10mm on the APS-C, 15mm for the 135 format camera) .. and I want the entire images from both cameras .. no cropping!
The advantage is now back with the D800 simply due to the resampling difference.

So, because of the resampling factor, I've assessed that the D500 would make a better low light/sports/birding type camera and the D800 would still be my preferred camera for landscapes.

And as for upsampling again .. whilst it can be done, I think it's probably best to go down with respect to sampling.
That is, if you need to compare, compare at the lower pixel count.
Upscaling is usually reserved for rare those times when you need a larger print than you can physically do.

[ameerat42]

Going up doesn't help much I'd imagine. If just for displaying on the interweb it might not make a lot of difference, if you're pushing extremes it might.
I don't really have a clue but thought I'd say something to get an automated thanks. I've never had one of them before (I think?);)

Thanks for this useful post, Mark..
Click on My Profile, then on Post Thanks.:nod:

--And see! You've got a bonus mechanical one.:D

[martycon]

thanks for the chuckle.