24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Started Jun 19, 2012 | Discussions
Jefftan Senior Member • Posts: 1,862
24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

I often see comparison done with downsampling 24MP to 16MP
should the comparison be better done upscaling 16MP to 24MP?

What is the best engine use for up or downscaling?
Lanczons 3 ?

Robsphoto
Robsphoto Senior Member • Posts: 1,218
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Jefftan wrote:

I often see comparison done with downsampling 24MP to 16MP
should the comparison be better done upscaling 16MP to 24MP?

I can fully appreciate the theoretical merits in downscaling and upscaling, and if I had to make a choice, I would say downscaling is my preference.

But, I can also see considerable merit in making comparisons without any upscaling or downscaling taking place, as explained here:

http://forums.dpreview.com/forums/read.asp?forum=1037&message=41818823

I think comparisons between 16mp and 24mp sensors should be a two-part process, one using upscaling or downscaling, and then making comparisons of the full-sized unscaled images from both sensors.

Cheers
Rob
http://www.robsphotography.co.nz/36-mp-FF-vs-24-mp-aps-c.html
Comparison of a 36mp FF camera with a 24mp APS-C camera such as Sony A77
The A77 pixels are much smaller than those of a 36mp FF camera!

philbot Contributing Member • Posts: 846
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Ultimately it shouldn't matter too much if assessing noise, both will theoretically introduce minuscule artefacts, it's just a vehicle to allow easy comparison by eye, if you have to nitpick at the level of artefacting, you can consider them as good as equal..

However, I think upscaling is probably the better option, since it guarantees you loose no detail from either sensor.. downscaling an ISO100 shot would certainly cause a loss in detail of the 24MP sensor.

But at the end of the day, our eyes/brain really can only make easy comparisons if they are viewed at the same 'size', they seem unable to look at images of different sizes and determine which has the most noise and detail overall unless it's massively obvious (the pixel-peepers dilema)..

OP Jefftan Senior Member • Posts: 1,862
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

I think the most important area of comparison is detail
whether 24MP sensor really has more detail at base ISO

if you look at those 18MP Sony point and shoot and compare it with 16 MP NEX-5N
Which has more detail?

Of course APS-C sensor is not the same but you get the idea. I believe 24 MP Sony APS-C is capturing more detail than 16 MP at base ISO

philbot Contributing Member • Posts: 846
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Jefftan wrote:

I think the most important area of comparison is detail
whether 24MP sensor really has more detail at base ISO

if you look at those 18MP Sony point and shoot and compare it with 16 MP NEX-5N
Which has more detail?

Of course APS-C sensor is not the same but you get the idea. I believe 24 MP Sony APS-C is capturing more detail than 16 MP at base ISO

Yes, the resolution (lph) tests show this is the case.. it's then a sliding scale as ISO increases, at some point they tend towards the same resolution since it's then all about sensor size, not pixel size.

zefelder Regular Member • Posts: 191
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Downscaling for high iso makes more practical sense. At ISO 6400 I would not expect 16mp sensor photos to be upscaled for printing and I'd expect 24mp photos to be normally downscaled. I'd actually expect the 16mp photos to be normally downscaled too at this ISO.

Regular scaling algorithms tend to damage image a bit. For example in my thread ( http://forums.dpreview.com/forums/readflat.asp?forum=1037&thread=41792634 ) test E was actually the test C image (from A55) upscaled & downscaled and people said it looked worse than C. In two cases upscaled & downscaled C was ranked below downscaled D (A77 photo). This means that scaling photo causes IQ degradation at least comparable to IQ difference between 16mp and 24mp sensors.

Advanced scaling algorithms on the contrary use techniques to visually enhance the image. So whatever scaling algorithm is used for a (even) more scientific test both images should be downscaled (or upscaled) to the same size which is either 24mp.

That saying I don't know what such test is going to achieve as for anyone who base their opinion on facts, the practical equivalence of 16mp and 24mp sensors should be obvious from the recent comparison tests, and as for others - nothing will convince them, so why waste time?

cgarrard
cgarrard Forum Pro • Posts: 15,317
How about just making a big print

with a shot of the same scene, same exposure, etc.

Pick your print size up to 24x36, the larger the better.

C

http://photographic-central.blogspot.com/

Ironic Misquote: "Nothing can stop the man with the right mental attitude; nothing can help the man with wrong one."
-Thomas Jefferson

rinkos Senior Member • Posts: 2,227
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Jefftan wrote:

I often see comparison done with downsampling 24MP to 16MP
should the comparison be better done upscaling 16MP to 24MP?

What is the best engine use for up or downscaling?
Lanczons 3 ?

frankly non of those engines are good enough..there are special enlargment softwares our there which do a lot better then those resizing algo's

but i do agree ...the proper way should be upscaling as in downscaling details are lost ...nice catch indeed.

either way like i said about the special enlargement apps...many of those can enlarge a shot up to 4 times atleast without u noticing any difference.

Meshuggah Contributing Member • Posts: 511
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

philbot wrote:

Ultimately it shouldn't matter too much if assessing noise, both will theoretically introduce minuscule artefacts, it's just a vehicle to allow easy comparison by eye, if you have to nitpick at the level of artefacting, you can consider them as good as equal..

However, I think upscaling is probably the better option, since it guarantees you loose no detail from either sensor.. downscaling an ISO100 shot would certainly cause a loss in detail of the 24MP sensor.

But at the end of the day, our eyes/brain really can only make easy comparisons if they are viewed at the same 'size', they seem unable to look at images of different sizes and determine which has the most noise and detail overall unless it's massively obvious (the pixel-peepers dilema)..

perhaps this is true wit some, not with others. what is really happening is some of us are satisfied with the equality of two objects similarity, bc there are many things that get compared and are not the same size. technically, almost nothing in the natural world is exactly the same size, so it really comes down to when you can compare different size objects accurately. some may be better at it than others, some may need it to be more similar than others.

i do agree upscaling is a more accurate comparison. this will not reduce 24mp noise but will also possibly soften the lower MP file. either way, it makes more sense as the whole purpose for a larger rez is to allow things like cropping. it makes no sense to downsample for any reason, other than for web use ect.

so lets see some of these same comparisons with upscaling now. if sony can introduce a decent feature in their smart zoom, id imagin PP will do an even better job. if i had two different rez cameras in my lap right now, id be all over the upscaling threads.

TrojMacReady
TrojMacReady Veteran Member • Posts: 8,722
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Meshuggah wrote:

philbot wrote:

Ultimately it shouldn't matter too much if assessing noise, both will theoretically introduce minuscule artefacts, it's just a vehicle to allow easy comparison by eye, if you have to nitpick at the level of artefacting, you can consider them as good as equal..

However, I think upscaling is probably the better option, since it guarantees you loose no detail from either sensor.. downscaling an ISO100 shot would certainly cause a loss in detail of the 24MP sensor.

But at the end of the day, our eyes/brain really can only make easy comparisons if they are viewed at the same 'size', they seem unable to look at images of different sizes and determine which has the most noise and detail overall unless it's massively obvious (the pixel-peepers dilema)..

perhaps this is true wit some, not with others. what is really happening is some of us are satisfied with the equality of two objects similarity, bc there are many things that get compared and are not the same size. technically, almost nothing in the natural world is exactly the same size, so it really comes down to when you can compare different size objects accurately. some may be better at it than others, some may need it to be more similar than others.

i do agree upscaling is a more accurate comparison. this will not reduce 24mp noise but will also possibly soften the lower MP file. either way, it makes more sense as the whole purpose for a larger rez is to allow things like cropping. it makes no sense to downsample for any reason, other than for web use ect.

so lets see some of these same comparisons with upscaling now. if sony can introduce a decent feature in their smart zoom, id imagin PP will do an even better job. if i had two different rez cameras in my lap right now, id be all over the upscaling threads.

It's been done countless times. Pay attention.

123Mike Veteran Member • Posts: 4,643
Better than Lanczons 3 ? What's it called ?

Lanczons 3 ?

frankly non of those engines are good enough..there are special enlargment softwares our there which do a lot better then those resizing algo's

I'd like to know the name of the algorithm that is supposedly better.

D Williams Contributing Member • Posts: 804
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Meshuggah wrote:

... makes no sense to downsample for any reason, other than for web use ect.

'Web use and .etc' covers probably 99% of photos that are printed or viewed on a display.

 D Williams's gear list:D Williams's gear list
Panasonic Lumix DMC-LX7 Sony Alpha DSLR-A900 Sony Alpha NEX-7 Sony SLT-A77 Sony a6000 +30 more
mick232 Contributing Member • Posts: 848
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Jefftan wrote:

I often see comparison done with downsampling 24MP to 16MP
should the comparison be better done upscaling 16MP to 24MP?

Do you want to compare scaling algorithms or cameras?

 mick232's gear list:mick232's gear list
Sony Alpha NEX-5 Sony Alpha a99 Tamron SP AF 70-200mm F/2.8 Di LD (IF) MACRO Sony 75-300mm F4.5-5.6 Sony 135mm F1.8 ZA Carl Zeiss Sonnar T* +19 more
Dave Oddie Veteran Member • Posts: 4,075
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

rinkos wrote:

Jefftan wrote:

but i do agree ...the proper way should be upscaling as in downscaling details are lost ...nice catch indeed.

Why?

The recent test where the 24mp images were downscaled was to see how they compared to 16mp images that obviously hadn't been downscaled. As the tests showed there was nothing in it, any detail loss was irrelevant.

So printing either at 16x11 which gives roughly 300 dpi you would not be able to tell them apart.

Whether you would print ISO 6400 images at 16x11 is another matter!

To quote zefelder directly from his post in this thread:

"Downscaling for high iso makes more practical sense. At ISO 6400 I would not expect 16mp sensor photos to be upscaled for printing and I'd expect 24mp photos to be normally downscaled. I'd actually expect the 16mp photos to be normally downscaled too at this ISO."

This is true. You would normally be downscaling the photos anyway not printing them at 20 inches across.

 Dave Oddie's gear list:Dave Oddie's gear list
Sony SLT-A77 Sony DT 11-18mm F4.5-5.6 Sony DT 16-80mm F3.5-4.5 ZA Carl Zeiss Vario-Sonnar T* Sony 70-300mm F4.5-5.6 G SSM Sony 500mm F8 Reflex +4 more
OP Jefftan Senior Member • Posts: 1,862
Re: Better than Lanczons 3 ? What's it called ?

just use FSviewer and this Lanczons 3 is the default algorithm

There are about 10 not sure which is better

123Mike wrote:

Lanczons 3 ?

frankly non of those engines are good enough..there are special enlargment softwares our there which do a lot better then those resizing algo's

I'd like to know the name of the algorithm that is supposedly better.

OP Jefftan Senior Member • Posts: 1,862
I have done a new 24 vs 16MP comparison

in a new thread

http://forums.dpreview.com/forums/readflat.asp?forum=1037&message=41833790

Both pictures are form dpreview comparison sample at base ISO

NEX 7 downscale using fsviewer with default algorithm

As I can tell, NEX 7 indeed capture more detail but only slightly. You really have to pixel peep to see it

So basically at base ISO NEX 7 is slightly better

I haven't done ISO 3200 comparison yet will do so later

Dennis Forum Pro • Posts: 17,355
Re: 24MP vs 16MP should the proper way be upscaling 16 to 24 rather than reverse?

Jefftan wrote:

I believe 24 MP Sony APS-C is capturing more detail than 16 MP at base ISO

You can find tests, from dpreview to imaging-resource to Pop Photo that all state that this is true. What exactly are you hoping to prove with another test ? Personally, I think that if the incremental gain in detail you'd get going from 16MP APS-C to 24MP is really that important, I'd probably want to test it via printing, not trying to evaluate on a computer monitor.

  • Dennis

-- hide signature --
Keyboard shortcuts:
FForum MMy threads