Standard T2i vs High Resolution T2i

ray_parkhurst · Post by **ray_parkhurst** » Mon Jan 06, 2020 10:17 pm

I received a standard T2i today and tested it against my HRT2i with AA filter removed. Test subject is...a 1954-S Lincoln Cent. Critical focus was done on a middle-height feature, Lincoln's throat, which in the case of this coin has a blotch of corrosion. I used a SN8000ED set at ~0.7x magnification. I'm finding that this lens is near optimum one way at 0.7x, and 1.4x the other, so it's an excellent lens for Lincoln Cents on APS-C.

I'll show the overall view of the coin, and then center crops around the corrosion spot. I set sharpening to 0 and all other camera parameters the same on both cameras. What I saw before was a significant increase in false colors, but now I am seeing more sharpness improvement than I had remembered. Let me know what you all see...

Overall image showing area at throat with corrosion spot:

200% Crop, Standard T2i

200% Crop, HRT2i

RobertOToole · Post by **RobertOToole** » Mon Jan 06, 2020 10:58 pm

Hi Ray,

Definitely sharper no doubt, pretty easy to see.

Best,

Scarodactyl · Post by **Scarodactyl** » Tue Jan 07, 2020 12:33 am

I'm really impressed by the difference here. No wonder people are upset about aa filters.

rjlittlefield · Post by **rjlittlefield** » Tue Jan 07, 2020 1:08 am

Were the T2i and HRT2i images processed optimally in each case?

If not, then of course we're going to see a lot more sharpness without the AA filter. But that's not a comparison that I'm particularly interested in. What I care about is final image quality, after relevant tuning to correct for AA blurring if appropriate.

Tuned for same sharpness, that is, same system MTF, I'd expect some extra noise in the AA case because stronger sharpening would be needed to back out the AA blurring. But offhand I can't predict how serious that would be.

I am reminded of https://www.photomacrography.net/forum/ ... 535#104535 .

--Rik

Macrero · Post by **Macrero** » Tue Jan 07, 2020 3:37 am

Ray,

Thank you for taking the time to make and share this comparison.

The difference is obvious, as was to be expected.

It would be interesting to see the comparison that Rik suggests.

In my experience you can pretty much match the level of sharpness of the AA-less image, but you can't recover some fine detail that was not captured.

Furthermore, the extra sharpening/deconvolution required by the anti-aliased file results in noise increase which is much more pronounced with small sensors with poor SNR. And also the risk of sharpening artifacts and/or halos is higher.

In summary: Anti-aliased images are trickier to process and to end up with a clean, sharp, detailed result, than an AA-less ones.

Best,

- Macrero

Adalbert · Post by **Adalbert** » Tue Jan 07, 2020 4:25 am

Hi Ray,
Thank you for the test!
Unfortunately I cannot see any difference of the resolution.
Could you please take the photographs of the test target or scales?
Thank you in advance.
BR, ADi

chris_ma · Post by **chris_ma** » Tue Jan 07, 2020 5:05 am

in my experience the difference depends a lot on the subject.
a sensor without low-pass filter glass will of course have more detail, but also more false color detail/aliasing on very fine structures.
so a part what is thought of being extra resolution is just fake structures coming from the debayer algorithm. here is an example with the M6II to some cameras without OLPF, and one with pixel shift which reveals the real image structure:
link

[edit: tried to shorten the ugly link]

Macrero · Post by **Macrero** » Tue Jan 07, 2020 5:47 am

chris_ma wrote:in my experience the difference depends a lot on the subject.
a sensor without low-pass filter glass will of course have more detail, but also more false color detail/aliasing on very fine structures.
so a part what is thought of being extra resolution is just fake structures coming from the debayer algorithm. here is an example with the M6II to some cameras without OLPF, and one with pixel shift which reveals the real image structure:

Here is a crop of the 5DS R RAW with better demosaicing:

https://images2.imgbox.com/31/b3/y5ok9vlw_o.jpg

yep, there is still false detail, but looks better than the DPR sample.

A perfect AA filter should result in capturing all the real detail that can be captured, discarding the false detail and avoiding aliasing/moire. But no AA filter is perfect, and you actually often lose real detail as well.

So the question is: Sharp image with a mix of real and false detail or a blurred one with a poor real detail? I choose the first option.

- Macrero

EDIT: Comparing apples and oranges here, these was not meant to be a comparison, but FWIW, here are a couple of 100% crops of stacks of the same fly, one with a Canon M6 II (AA) and other with a Fuji X-T1 (AA-less).

Both shot with a Lomo 3.7 + Rodagon 150. The Canon image was stronger pre-sharpened in-RAW.

Canon:

https://images2.imgbox.com/96/75/yKeMiE0D_o.jpg

Fuji:

https://images2.imgbox.com/c5/81/8jHf4hOj_o.jpg

ray_parkhurst · Post by **ray_parkhurst** » Tue Jan 07, 2020 6:23 am

All...I made a typo in my OP. The crops are presented at 200%, not 100%.

As I stated in a previous post, my more recent interest in pixel shifting and SR is making the lack of AA filter seem much more critical. My tests last night were just a quick check, but I do plan to reproduce them with SR to see if I'm right about my expectations. Stay tuned...

chris_ma · Post by **chris_ma** » Tue Jan 07, 2020 7:51 am

Macrero wrote:Here is a crop of the 5DS R RAW with better demosaicing:
https://images2.imgbox.com/31/b3/y5ok9vlw_o.jpg
yep, there is still false detail, but looks better than the DPR sample.

hmmm, certainly has a lot less false color and looks sharper, but I'm not convinced about "better".

look at the face of the woman to the left and compare it with the S1R PS capture - in your example there are a lot of sharp detailed vertical lines there while the original face consists of only horizontal lines.
so all this apparently sharp detail is totally false.
same thing happens in the left wooden leg of the picture stand, only this time long vertical lines suddenly turned into short horizontal ones..

btw, I agree that downloading the RAW can give a bit more quality then the online tool, but it's very convenient to do a quick first comparison.

Macrero · Post by **Macrero** » Tue Jan 07, 2020 8:34 am

chris_ma wrote:hmmm, certainly has a lot less false color and looks sharper, but I'm not convinced about "better".

look at the face of the woman to the left and compare it with the S1R PS capture - in your example there are a lot of sharp detailed vertical lines there while the original face consists of only horizontal lines.
so all this apparently sharp detail is totally false.
same thing happens in the left wooden leg of the picture stand, only this time long vertical lines suddenly turned into short horizontal ones..

btw, I agree that downloading the RAW can give a bit more quality then the online tool, but it's very convenient to do a quick first comparison.

Well, there are parts that looks worse (sharper, but with false detail) and others that looks better (sharper and with more real detail).

https://images2.imgbox.com/5b/00/JyD04CoY_o.png

There will be always a penalty. The perfect sensor does not exist. Not even the S1-R pixel shifted image is "perfect". I still prefer the sharper image. In real life shooting the difference between the potential false detail in some zones of the image and the real detail is hard to tell. And unless we talk critical, scientific work, it would not matter.

Look at the M6 II and X-T1 crops I posted. You see obvious false detail in the Fuji's one? I do not. But it looks much sharper and clearer.

- Macrero

ray_parkhurst · Post by **ray_parkhurst** » Tue Jan 07, 2020 8:37 am

I shot the coin again with both cameras, taking 8 shots on a diagonal with ~5um steps, and then ran SR on both. See results below. I'm seeing significantly more detail and sharpness in both cameras, but camera-camera comparison is still similar. 200% crops again.

T2i with AA filter:

T2i without AA filter:

RobertOToole · Post by **RobertOToole** » Tue Jan 07, 2020 9:22 am

Sensor stack cover glass: thick or thin?

Recently I spent some time looking into the subject of sensor stack glass thickness when I was setting up my 1x test.

Thinner sensor stack cover glass would be more of an advantage in sharpness a lens designed without one, like a printing-nikkor, but the thinner cover glass could have a negative effect on a newer lens designed to work with a thick sensor stack like a Sony 90G Oss that is made to work with a 3mm slab of glass in place on a Sony sensor.

Sensor stack thickness can also effect chromatic aberrations, especially in the corners.

I included links on the subject at the bottom of the 1x test on my site. I can post these if anyone is interested. There is quite a bit of info about this online.

Best,

Robert

ray_parkhurst · Post by **ray_parkhurst** » Tue Jan 07, 2020 9:48 am

According to the folks who did my conversion, when they remove the AA filter it is replaced by a piece of glass the same thickness. That said, the magnification changes between the with/without filter T2i's, as you can see in the pics. I only swapped the cameras for the comparison, keeping the same bellows extension, yet had to refocus between shots, so I assume the AA filter has more effect than just its thickness.

Edited to add: the process I used was to downsize the resulting composited image by 2x, then upscale it by 2x to present.

Adalbert · Post by **Adalbert** » Tue Jan 07, 2020 10:13 am

Hello Macrero, hello Chris,
Therefore I have suggested the scales :-)
Please take a look at: https://www.photomacrography.net/forum/ ... highlight=
BR, ADi