Pages: [1]   Go Down

Author Topic: Do parallax corrected subframes *need* to be warped when pano stitching?  (Read 1636 times)

capital

  • Full Member
  • ***
  • Offline Offline
  • Posts: 222
    • Website

Do parallax corrected subframes *need* to be warped when pano stitching? That is, can a pano comprised of images revolved around the nodal point be successfully stitched by repositioning only or do they need to be warped to some projection? My concern is pixel level smearing due to warping. My guess is that if the overlap is great enough, then the geometric distortion of the lens will be small enough to neglect the need for warping.
Logged

kers

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4388
    • Pieter Kers
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #1 on: February 23, 2015, 04:08:36 am »

....My concern is pixel level smearing due to warping. ....

Hello, i think yes- in almost all circumstances the warping is necessary.

If your concern is pixel smearing on pixel level - it always will happen if you reposition images that have some distortion...
I have only one lens without distortion (almost)...and even then the change of the angle changes the perspective and that also has to be compensated....( in the case of a rectangular image result- normal picture)

The exception is if you use a TS-lens - now you can use the shift and repositioning is well possible - the lens (entrance pupil) should then be static- and the camera is doing the shift

cheers PK
« Last Edit: February 23, 2015, 04:10:39 am by kers »
Logged
Pieter Kers
www.beeld.nu/la

Petrus

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 952
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #2 on: February 23, 2015, 04:19:29 am »

All lenses, no matter how perfect they are, introduce some elongation at the corners. The shorter the focal length the more of this happens. It is not lens distortion, it is an unavoidable fact of rectilinear projection. In two adjacent frames this elongation goes the opposite ways in the adjoining side of the pictures. This makes it necessary to warp the images to make them match.

This process (for single row panoramas) usually converts the projection from rectilinear to cylindrical. For longer panoramas this is the only possibility, because already a 180 degree panorama stitched to rectilinear projection would stretch to infinity on both ends. Highly unpractical at least for prints...
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #3 on: February 23, 2015, 04:36:00 am »

Do parallax corrected subframes *need* to be warped when pano stitching? That is, can a pano comprised of images revolved around the nodal point be successfully stitched by repositioning only or do they need to be warped to some projection?

Hi,

The short answer is, yes. The longer answer is, most likely yes, unless you want to produce output with flat panels that are arranged in a spherical orientation.

Whenever you want to project an image on a surface, the projection will distort the image. Take for instance the left most image of a single row of three images. The camera, presumably with a rectilinear lens and a flat sensor, looked at that part of the scene at an angle, different from the center image. That angle distorted the image relative to the center image, and thus needs reprojection if you want it to not produce abrupt kinks in straight lines and edges. Same for the right most image but in opposite direction.

Add to this that no lens is absolutely free of distortions, and even perfect perpendicular alignment of the optical axis on the middle of the sensor is within a small margin of error. The stitching software can take care of all of those variables.

Quote
My concern is pixel level smearing due to warping.

Dedicated stitching software typically offers a choice of resampling algorithms, and they are usually a lot better than Photoshop's Bicubic in maintaining micro-contrast, even with sub-pixel shifts. Shooting panoramas, is also usually done with a longer focal length than one would otherwise use to get the FOV. Longer focal lenses are usually better corrected, especially when dealing with edge and corner detail, and they offer higher resolution due to a larger magnification on sensor. So you build in additional resolution which can partly be used by resampling.

In addition, many panoramas are so large that they can be down-sampled without sacrificing output resolution, or at least they need less upsampling, which all benefits final resolution.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

capital

  • Full Member
  • ***
  • Offline Offline
  • Posts: 222
    • Website
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #4 on: February 23, 2015, 05:10:45 am »

Thank you Kers, Petrus & Bart for your timely replies. Okay, so a transformation is needed to be "correct" but is there any IQ advantage (from warping) of combining 10 overlapping sub frames, versus say 3 overlapping sub frames, if 3 sub frames is the minimum needed to cover the complete frame?

In case anyone is wondering why I am concerned about the pixel level smearing is that I am using a 4.7 MP X3F sensor for the stitch, so every pixel counts.
« Last Edit: February 23, 2015, 05:14:23 am by capital »
Logged

Petrus

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 952
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #5 on: February 23, 2015, 05:34:12 am »

Thank you Kers, Petrus & Bart for your timely replies. Okay, so a transformation is needed to be "correct" but is there any IQ advantage (from warping) of combining 10 overlapping sub frames, versus say 3 overlapping sub frames, if 3 sub frames is the minimum needed to cover the complete frame?


There is, as less warping is needed in each case. For this reason the "best practice" is to stitch many vertical frames instead of few horizontal ones.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #6 on: February 23, 2015, 05:57:57 am »

Thank you Kers, Petrus & Bart for your timely replies. Okay, so a transformation is needed to be "correct" but is there any IQ advantage (from warping) of combining 10 overlapping sub frames, versus say 3 overlapping sub frames, if 3 sub frames is the minimum needed to cover the complete frame?

As 'Petrus' said, the common approach (assuming a landscape oriented pano, i.e. wider than high) is to shoot tiles with the camera in portrait orientation (that will also reduce the number of rows required). You'll need a decent overlap between image tiles, say between 20% and 50% to be able and tackle difficult situations, e.g. where subject motion is involved, or significant shifts in brightness or color temperature. Using the wider overlaps reduces the required amount of projection distortion between images a bit (especially on projection types like cylindrical).

Quote
In case anyone is wondering why I am concerned about the pixel level smearing is that I am using a 4.7 MP X3F sensor for the stitch, so every pixel counts.

I see, but the principle remains the same. Use a longer focal length lens if you can change it (and adjust aperture for DOF), and therefore also generate more tiles, with more additional resolution than you are likely to lose with good resampling quality. Also understand that very wide panos may require a different type of output projection than rectilinear. But if you stitch for the purpose of gaining resolution, rectilinear projection is usually not an issue because that's how most lenses work to begin with (on a flat sensor and for flat output).

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Petrus

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 952
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #7 on: February 23, 2015, 06:44:47 am »

Also understand that very wide panos may require a different type of output projection than rectilinear.

"Very wide" does not necessarily mean more than 100 degrees or so. Corner elongation starts to be quite bad after that, and it is the same for supewide lens and a rectilinear panorama.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #8 on: February 23, 2015, 06:57:10 am »

"Very wide" does not necessarily mean more than 100 degrees or so. Corner elongation starts to be quite bad after that, and it is the same for supewide lens and a rectilinear panorama.

Beyond a Field of View of 100 to 120 degrees, the necessary stretching/warping of pixels will also affect per pixel resolution. That is exactly what the OP wants to avoid. The problem is not the required warping per se, but the fact that we want to view such images from the wrong viewpoint. If we watch the image from the correct position, which is uncomfortably close, the stretched pixels will look completely undistorted, and resolution is restored by perspective that compresses features to how they originally were.

It's just like looking at text painted on the road surface, when viewed from the intended viewpoint, it looks perfectly legible. But when looked at from the wrong viewpoint, it looks stretched. However, we need to accept that most very wide panos are viewed from too far a distance. It's all about projection distortion.

Cheers,
Bart
« Last Edit: February 23, 2015, 05:43:08 pm by BartvanderWolf »
Logged
== If you do what you did, you'll get what you got. ==

capital

  • Full Member
  • ***
  • Offline Offline
  • Posts: 222
    • Website
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #9 on: February 23, 2015, 03:09:43 pm »

Thank you Bart & Petrus for the continued discussions.
Logged

kers

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4388
    • Pieter Kers
Re: Do parallax corrected subframes *need* to be warped when pano stitching?
« Reply #10 on: February 26, 2015, 03:47:44 am »

Thank you Kers, Petrus & Bart for your timely replies. Okay, so a transformation is needed to be "correct" but is there any IQ advantage (from warping) of combining 10 overlapping sub frames, versus say 3 overlapping sub frames, if 3 sub frames is the minimum needed to cover the complete frame?
In case anyone is wondering why I am concerned about the pixel level smearing is that I am using a 4.7 MP X3F sensor for the stitch, so every pixel counts.
I can add that if you use 20 images to make the same panorama ( with a telelens) instead of 10 (with a normal lens)
you will have simply twice the quality in the panorama with the telelens...
Logged
Pieter Kers
www.beeld.nu/la
Pages: [1]   Go Up