Super resolution with autopano?

[psmeraldi]

Super resolution with autopano?

by psmeraldi » Thu Jan 01, 2015 9:20 pm

Hi, I have read this tutorial about reducing noise with autopano.

http://www.kolor.com/wiki-en/action/vie ... ll_Sensors

I know that there are similar techniques to increase resolution; usually they are called "super resolution" and they are used in astrophotography.

Is it possible to do something similiar with autopano? Like taking a series of picture of a static subject at the maximum focal lenght available and stacking them to recover some detail?

Here are some details

https://aperture64.wordpress.com/2014/0 ... esolution/

Thanks

Paolo

[HansKeesom]

Re: Super resolution with autopano?

by HansKeesom » Fri Jan 02, 2015 1:02 am

Instead of making one photo with a certain lens, you can shoot a number of rows and columns with a longer lens and then stitch them together using autopano

I process and stitch for a number of photographers. Some make 180 degree shots using a 50 MM lens by shooting 18 photos a row and 4, 5 or more rows. The resulting files are realy highres.

the same way I did a panorama of a beach that is now on a 11 m wall printed at 150 dpi quality, talking about superresolution......

[gkaefer]

Re: Super resolution with autopano?

by gkaefer » Fri Jan 02, 2015 2:08 pm

i use oloneo to stack the images for denoising them:
http://www.oloneo.com/en/page/products/ ... noise.html
the resulting images I do stitch with apg.
if the fous is on focusstacking than I use zerene stacker:
http://zerenesystems.com/cms/stacker
the advantage here is that apg can handle the unstacked source files using zerenstacker via builtin plugin, so finally the images get stacked automatically with zerene and the resulting images stitched with apg...
Georg

[klausesser]

Re: Super resolution with autopano?

by klausesser » Fri Jan 02, 2015 4:24 pm

Instead of making one photo with a certain lens, you can shoot a number of rows and columns with a longer lens and then stitch them together using autopano

The goal is to reduce noise to enhance the resolution - not to gain resolution by stitching.

Works well! Oloneo, Helicon, FocusMagic, ZereneStacker can do that.

Have a very happy and wealthy 2015 alltogether!

Klaus

[HansKeesom]

Re: Super resolution with autopano?

by HansKeesom » Sat Jan 03, 2015 10:06 pm

The question I answered to was "Is it possible to do something similiar with autopano?"

[psmeraldi]

Re: Super resolution with autopano?

by psmeraldi » Sat Jan 03, 2015 11:05 pm

Hanskeesom, thanks for your reply, but what you describe here (photo stitching) works as well as the longest focal lenght you have.
Insteas of taking a single photo with a wide angle lens, you can take ten pictures with a telephoto lens, or 100 with a super tele and of course you have more resolution.

The point is, if you have a 300mm lens, can you get more resolution of the same angle of view, that is, can you stack photos and get the same resolutions as if you had a 400 or 500?
With stitching, you can get more resolution with a 300 on the same area a 200 would frame, but not if compared to a 400 or 500.


The process is described here

http://photoacute.com/tech/superresolution_faq.html

It seems to me that autopano could be tweaked or used to this purpose.

Thanks Georg and Klaus, I did not know these software.

[HansKeesom]

Re: Super resolution with autopano?

by HansKeesom » Sun Jan 04, 2015 1:13 am

If you limit yourself to the situation you describe, you might want to look at the size % of the rendering. Since recently it goes all the way up to 150, not sure what that means.
In general I don't think autopano can do superresolution as you describe it.

[klausesser]

Re: Super resolution with autopano?

by klausesser » Sun Jan 04, 2015 2:00 am

Thanks Georg and Klaus, I did not know these software.

Hi!
Let´s talk about "superresolution" and REAL high resolution.

"Superresolution" is NOT the same as high resolution - it describes an "enhanced" normal resolution
the amount your camera can provide with one shot.

Real high resolution you ONLY get by by gathering very much pixels using a very good lens.

You can use a longer than wide-lens for achieving a wide field of view (fov) AND very high resolution.
An example: imagine you want to do a wide-angle shot with a fov like 20mm superwide angle lens.
You can use a 20mm lens and a D800 camera provides 36mpx.
You can use a 50mm lens and the D800 provids a MUCH higher resolution by taking 6 or 8 shots spinning around
the NPP (NoParallaxPoint) and stitching the shots for a planar projection.

Achieving this wider fov you also achieve multiple numbers of pixels for getting a REAL high resolution.

Indeed this way definitely is preferable when you aim to do big prints or achieving deep zooms on-screen.

In this regard Hans was completely right. I also would suggest the way of stitching several shots using a NPP adapter
like a NodalNinja or others.

best, Klaus

[HansKeesom]

Re: Super resolution with autopano?

by HansKeesom » Sun Jan 04, 2015 11:01 am

I do the stitching for a number of professional photographers who don't have the time, big enough computer, knowledge or software for it.
They often use the 50 mm to make a 180 degrees landscape picture. Takes them (only) half an hour to take all the photos (brackets of 5 or 7).
There is so much detail in the end result. Indeed, great for big prints.

[klausesser]

Re: Super resolution with autopano?

by klausesser » Sun Jan 04, 2015 5:39 pm

It seems to me that autopano could be tweaked or used to this purpose.

Thanks Georg and Klaus, I did not know these software.

Using Autopano absolutely makes no sense using it this way as long as you do not SHOOT correctly
using a NPP adapter and STITCH the images.

What you call "superresolution" is something different and - i repeat - does NOT increase your resolution
in terms of a bigger amount of pixels . . which is the only way to REALLY gain a HIGH resolution.

"Superresolution" is a somewhat "stained" euphemism for achieving a bit more clarity in the usual
way of shooting.
It was indeed used in astro-photography for showing *clearer* details - but not by showing them bigger,
as it is when you shoot for stitching images by using longer lenses.
This also can be done in astro-photography - but that´s a bit tricky.

Klaus

[slipstick]

Re: Super resolution with autopano?

by slipstick » Mon Jan 05, 2015 12:04 am

What you call "superresolution" is something different and - i repeat - does NOT increase your resolution
in terms of a bigger amount of pixels

That isn't correct. Have a look at the link previously provided: http://photoacute.com/tech/superresolution_faq.html

The resulting images are four times the size: twice the number of pixels on length and width. And it is "real", not interpolated, resolution, building on the additional information each image in a stack brings.

The downside is the requirement for individual lens / camera combination configurations. So one has to shoot with a combination that is known the photoacute for best results.

[klausesser]

Re: Super resolution with autopano?

by klausesser » Mon Jan 05, 2015 12:37 am

What you call "superresolution" is something different and - i repeat - does NOT increase your resolution
in terms of a bigger amount of pixels

That isn't correct. Have a look at the link previously provided: http://photoacute.com/tech/superresolution_faq.html

The resulting images are four times the size: twice the number of pixels on length and width. And it is "real", not interpolated, resolution, building on the additional information each image in a stack brings.

The downside is the requirement for individual lens / camera combination configurations. So one has to shoot with a combination that is known the photoacute for best results.

You´re getting something wrong here. A real gain in resolution will only take place when you double or quadruple the physical amount of pixels. This does not happen this way.
Without stitching it only happens when the sensor shifts pixel-wise in x and one in y direction. This way the Hasselblad 200MPx back does work.

The theory behind "superresolution" is different. I´m using PhotoAcute for years as a focus-stacker - and of course i also tested it´s "superresolution" . . .


But i´m not here to lecture you

best, Klaus

[slipstick]

Re: Super resolution with autopano?

by slipstick » Mon Jan 05, 2015 2:22 am

[...]A real gain in resolution will only take place when you double or quadruple the physical amount of pixels.

http://photoacute.com/tech/superresolution_faq.html#q2

It's real.

[klausesser]

Re: Super resolution with autopano?

by klausesser » Mon Jan 05, 2015 3:06 am

[...]A real gain in resolution will only take place when you double or quadruple the physical amount of pixels.

http://photoacute.com/tech/superresolution_faq.html#q2

It's real.

The physical amount of pixels - which we call "resolution" - is the number of the sensors (sensels) on the chip which we call "the sensor".

The number of pixels of this sensor defines "the resolution".

When you mechanically shift the whole sensor in x and y direction you can double or quadruple the physical resolution.
This is very expensive - Hasselblad´s digital back which provides 200MPx is a 50MPx back by nature. It works by shifting the sensor for achieving 200MPx.

The theory of "multiresolution" is enhancing the resolution by using mathematical processes - basing on the idea that shooting multiple times the same image can produce a kind of pixel"shift" basing on microscopic movements of the camera between the shots.
Basically you have not 4 times the pixels at 100% the same place by camera-shake and other when you shoot 4 times without moving the camera.
This works indeed - but to a very much smaller amount than you seem to think.

Doubling the resolution this way simply is a somewhat ridiculous joke, sorry. As i experienced in praxi there is no remarkable gain in resolution at all (i anyway use excellent cameras and lenses, btw.)

On the other hand *stitching* REALLY provides MUCH larger resolution - the physical/mathematical reasons for that are very transparent and commonly well known.
Given you want to shoot a wide-angle image: you can do it by:
1) using a wide-angle lens shooting ONE image (or multiple images for "superresolution")
or you use a
2) "normal" lens or "tele" lens for achieving the same field of view and stitch the images in APG or PTGui.

This way you REALLY achieve a MUCH higher resolution than by using a wide-angle lens.

But - as we say here: facts tell more than 1000 words. So i ask you to simply do it.

Klaus

[slipstick]

Re: Super resolution with autopano?

by slipstick » Mon Jan 05, 2015 3:34 pm

Doubling the resolution this way simply is a somewhat ridiculous joke, sorry.

Well, a) I don't think you're sorry and b) you're incorrect. Characterizing photoacute's results as a "ridiculous joke" just doesn't stand up. Single-image interpolation (such as Lanczos resampling) has significant limitations (and yet still don't deserve such a pejorative as "ridiculous joke"), but the multi-image technique used by photoacute and similar software produces images that, to all appearances, have come from much higher pixel-count sensors. I've had good results from landscapes, for example, showing clear barn features that were just blobs on the original-resolution images. The examples on the website ( http://www.photoacute.com/studio/exampl ... index.html ) are representative of my results: better real resolution with less noise.

One could in theory run raws through photoacute and then stitch the results ( I might try that when winter turns into spring ), but it would be interesting to have the algorithm built-in to APG, in effect doubling resolution without buying a new lens and stitching additional images, and without resorting to multi-row stitches. Simpler workflow.

[HansKeesom]

Re: Super resolution with autopano?

by HansKeesom » Mon Jan 05, 2015 7:28 pm

I am not impressed with the examples. Actually I am even more convinced better results can be achieved by stitching 7 partly overlapping photos made with a longer lens.....
"Without buying a new lens" is never an excuse for lack of quality, teleconvertors are not that expensive btw.

My workflow is just getting simpler when photographers use longer lenses.... and 7 overlapping photos is not more work then 7 photos of the same composition.

Finally, try feed 7 photos of same composition into autopano, you might be surprised.

[klausesser]

Re: Super resolution with autopano?

by klausesser » Mon Jan 05, 2015 7:49 pm

Doubling the resolution this way simply is a somewhat ridiculous joke, sorry.

Well, a) I don't think you're sorry and b) you're incorrect. in effect doubling resolution without buying a new lens and stitching additional images, and without resorting to multi-row stitches. Simpler workflow.

1) You can ride from here to Wladiwostok on a bycicle. You really can do that
2) "Doubling the resolution" would mean: i can take my 36MPx camera, do some identical shots, import them into PhotoAcute - and suddenly i get 72MPx . . . would be GREAT!!
3) I have a feel - correct me if i´m wrong - you are somewhat new to digital photogragphy. Are you?
4) what do you mean: without buying a new lens"?? You don´t need any new lens for stitching - you can use any lens you already have.

See: I am what one might call a "resolution-freak"

Being a professional photographer for some decades i´m shooting digital since 1996 - when a professional 6MPx-camera cost about 50000.-DM (25000.-€), working with each and every digital format since then. Actually i use 24x36mm "fullframe"-format (Nikon D800 and D810) and 33.1 x 44.2 mm mid-format (Hasselblad HD4/50). I use the MF back also on my largeformat-cameras Linhof and Sinar 4x5" with a slide-adapter for enhancing resolution
by shifting the sensor around in x and y direction. Works fine and is used by many photographers in the studio.
Analogue i still use 4x5", and 8x10" - that´s appr. 9x12cm and 18x24cm.

I deal with high resolution commercially: http://www.360impressions.de/HiRez/ (have a close look)
that means 4 - 12 Gigapixels and more for very large prints of 20m width and more in excellent quality.

As said: being a "resolution freak" i constantly gather all informations can i find about optimizing the resolution
i need for my work.

Testing PhotoAcute was one of the experiments i did - and other professional photographers did it also.

Result: "doubling the resolution" of a camera isn´t possible this way - because it happens by mathematically processing multiple shots instead of adding physical pixels . . that´s logic.

PhotoAcute does this in a very clever way using "multishot". Hasselblad also uses "multishot" - but before each shot the whole sensor moves in x and y direction. THIS way they REALLY can double or quadruple the resolution: they do it physically.
"Multishot" in that case means: multiple exposures on the same frame by shifting the sensor each exposure.
That´s why you can use the back only for non-moving objects.

"Multishots" WITHOUT shifting the sensor - like PhotoAcute does - can "enhance" the shot´s resolution a bit, yes.
It maybe can gain some additional resolution by using micro-movements on the sensor or electrical differences
between the multiple exposures and processing them to a sharp image. Definitely it can minimize noise - which can be called "more resolution".

But in no way it can "double the resolution". For achieving double resolution the sensor´s movement would
need to be a) larger and b) absolutely controlled. That´s not what happens in usual cameras.

Klaus

[klausesser]

Re: Super resolution with autopano?

by klausesser » Mon Jan 05, 2015 8:11 pm

I am not impressed with the examples.

And they definitely do not demonstrate "doubling the resolution".

They show a bit more clearity in the shots and less noise. Ok - that definitely IS
more resolution.

I guess "slipstick" puzzles some terms here - and maybe so do i

"Clearer" photos showing less noise of course show more details = better resolution.

That´s like puting on new eyeglasses.

The term "doubling the resolution" usually means doubling the amount of pixels,

What´s happening here is a kind of "unsharp masking" - needing mutiple shots.

That results in a sharper imgage - but it doesn´t mean "doubling the resolution".

Klaus

Last edited by klausesser on Mon Jan 05, 2015 8:14 pm, edited 1 time in total.

[Nanard]

Re: Super resolution with autopano?

by Nanard » Tue Jan 06, 2015 3:13 pm

I think one should distinguish two aspects :
The mathematical one :
information cannot be 'invented', so by interpolation you can double the number of pixels but the amount of information, ie the number of original pixels does not change. Same by stacking N identical images, you cannot produce new information, the number of usefull pixels does not change.

The visual aspect is quite different, with two benefits :
. being random, the noise is different for each image of a stack and so it can be reduced
. images of a stack are never perfectly aligned, there are always micro misalignments, and with a mathematical process it's possible to increase the contrast. The visual sharpness is increased, as would be between a very bad lens and a very good one with the same sensor, the image is much better but the number of pixels is the same.
Bernard

[mediavets]

Re: Super resolution with autopano?

by mediavets » Tue Jan 06, 2015 3:31 pm

I think one should distinguish two aspects :
The mathematical one :
information cannot be 'invented', so by interpolation you can double the number of pixels but the amount of information, ie the number of original pixels does not change.
Bernard

Surely it's the case that interpolation does indeed invent new information when increasing the pixel dimensions of an image; the extra pixels are invented by the software.

[Nanard]

Re: Super resolution with autopano?

by Nanard » Tue Jan 06, 2015 3:38 pm

Surely it's the case that interpolation does indeed invent new information when increasing the pixel dimensions of an image; the extra pixels are invented by the software.

Interpolation produces extra, 'artificial', pixels, but does not produce new information in the mathematical sense.

[Nanard]

Re: Super resolution with autopano?

by Nanard » Tue Jan 06, 2015 4:06 pm

Surely it's the case that interpolation does indeed invent new information when increasing the pixel dimensions of an image; the extra pixels are invented by the software.

Start with a 12 Mpix image, by interpolation produce a 48 Mpix file, print it 3 meters long, what kind of quality do you get ??
Bernard

[HansKeesom]

Re: Super resolution with autopano?

by HansKeesom » Tue Jan 06, 2015 5:13 pm

We must remember the superresolution technology is not the same as interpolation.

Last year I produced a file for a 11*2.5 m print at resolution 150 dpi. I did use APG. No interpolation was used.
We initially had a lower resolution file which I interpolated from 50 dpi to 150 dpi but it was unacceptable for the printing company and rightly so. There is so much more detail in a true 150 dpi file compared to an "interpolated to 150 dpi" file.

What was interesting is that I shot the panorama three times, at 70 mm, 200 mm and 250 mm. APG nicely merged everything together.

[klausesser]

Re: Super resolution with autopano?

by klausesser » Tue Jan 06, 2015 11:11 pm

error

Last edited by klausesser on Tue Jan 06, 2015 11:20 pm, edited 1 time in total.

[klausesser]

Re: Super resolution with autopano?

by klausesser » Tue Jan 06, 2015 11:19 pm

I think one should distinguish two aspects :
The mathematical one :
information cannot be 'invented', so by interpolation you can double the number of pixels but the amount of information, ie the number of original pixels does not change.
Bernard

Surely it's the case that interpolation does indeed invent new information when increasing the pixel dimensions of an image; the extra pixels are invented by the software.

Well - but the "new" information which it "invents" is garbage

Interpolation=upscaling for that reason minimizes visual (real) resolution by smudging details.

Klaus