Very short focal fisheye lenses
and
Stitching/Blending Double-Shot (DS) images

Part (2)

Summary of the study

Sigma and Sunex have recently and almost simultaneously put on the market a new very short focal lens. Both the Sigma 8-mm f2.8 and the Sunex Superfisheye 5.6-mm f5.6 may allow to do spherical panographies by assembling only two hemispherical opposite images (aka Double Shot or DS) thanks to their Angle of View >>180 degrees.
The noteworthy point is that this can be done now on an affordable basis unlike with the previously available but expensive alternatives (4.88-mm f5.2 and 7.45-mm f2.8) sold by Coastal Optical (also known as IPIX lenses) and not quoting the venerable (220 deg FOV) Nikon 6-mm f2.8 monster.

Part 1
Appraisal of the lenses performance shall be done in a pure practical approach without Chart measurement: I did not have the chance to put my hands on the lenses and had only some kindly provided sample images for experimentation.
While the Sunex is a fixed aperture model, the Sigma can be set from f/2.8 through f/22 and the capability of this lens shall be evaluated by taking this feature into account. The illustrated results are presented in Part 1.

Part 2 (This page)
Evaluation of panorama stitching programs likely to allow DS stitching is presented in Part 2 (This page).

Part 3
In the course of evaluation of possible ways to produce panorama with DS, I have discovered an anomaly
I have thus elected to describe the conditions where a strange dysfunction of some blending engines happens seemingly on a random way, when they are faced to blending DS images.
After many hour of experimentation, I have yet not found a valid explanation for this instability and I have resigned myself to write Part 3 in order to give all means to replicate for further investigation by others. I hope to be able to remove this part as soon as the problem is solved.

Note: The source images shot with the Sigma 4.5-mm were all kindly provided by Roberto Gomez Torres.


Part two

Double Shot panorama: Stitching Software performance appraisal

Introduction and background

Like the Sunex 5.6-mm f5.6 Superfisheye but unlike most other lenses, the new Sigma 4.5-mm f2.8 produces full circular images on the sensor of APS-C DSLR. Differing from its rivals, the Sigma fisheye lens can be fully controlled (exposure and focus) from the Nikon, Sigma or Canon camera
This Sigma new offspring shall therefore be the specimen for all the experimentation that follows.

As the effective Angle of View is larger than 180 degrees and as the sharpness of the image is quite decent up to the edge as long as the aperture is set at less than f/8, it looked to me as a nice possible candidate to return to Double-Shot that is the core of low end real estate agent panography.

I have read in a commercial from Sigma that the radial optical compression of the fisheye lens image has been especially designed as to be possibly easily re-mapped for scientific experimentation or observation purposes.
I have verified that this point holds also for panographic stitcher programs: most can handle DS correctly, but not all. This makes that lens preferable to an other possible solution (Raynox DCR CF 185 Pro converter) the mapping of which is known to challenge the algorithms of most stitcher programs and is generally used with four shots for an decent output quality.


Summary table

I have made the same panorama from the same images with different stitching software.
Discarding the PT Stitcher 1.8 ephemeral automatic version and since Photovista and IPIX seem not be available or are unsupported anymore, Stitcher Unlimited-DS and Panoweaver are the two programs that were dedicated to auto-DS from their initial introduction in the market.

I have tested both AutoPanoPro 1.4 and Panoweaver 4 and 5 with a Demo version.

Two Panorama Tools front ends have been tested. these are PTGui and Hugin (open-source graphical user interface (GUI)) as the latest versions of them are available on both MacOsS and Windows.

PTGui
Hugin
Panoweaver
AutopanoPro
Stitcher
Stitching program
v 7.5

(Mac and

Windows)

v 7.5

(Mac and

Windows)

w/ Panorama Tools Warping

v 0.7

(Mac and Windows)

v 4.0

(Mac)

v 5.0

( Only available on Windows)

AutopanoPro 1.4

(Mac)

Unlimited

MacOS Windows

Not DS

DS

(Mac and Windows)

Number of hemispherical images to stitch
DS, 3 and 4 images
DS, 3 and 4 images
DS, 3 and 4 images
DS then 3 images
DS then 3 images
DS, 3 and 4 images
3 then 4 images
DS (of course)
Automatic CP detection and Images Alignment
PTGui Auto-detection mode failed for DS.

Autopano-SIFT-detection mode failed for DS

Auto-detection mode

Auto-detection mode failed for DS

bad result w/ 3 images

Auto-detection mode

Auto-detection mode failed!!!!

Manual mode not tried

Manual CP placement and alignment optimization
OK (*)
OK (**)
Not tested
Not tested
Not tested
Not tested
Blender

engine

(***)

PTGui blender
Yes
Yes
NA
NA
NA
NA
NA
NA
Smart
NA
NA
NA
NA
NA
NA
Yes
Yes
Smartblend
Yes

(including on Mac)

Yes

(including on Mac)

Yes

(Indirectly)

NA
Yes
Yes
Yes (Mac)
Yes (Mac)
Enblend
Yes
Yes
Yes
NA
NA
Not

tested

Not

tested

Not

tested

PS CS3 Auto-blend layers
Yes

(in two steps)

Yes

(in two steps)

Not

tested

Not

tested

Not

tested

Not

tested

Not

tested

Not

tested

Caption:

Yes = I have trialed this configuration whatever its conclusion (successful, partly successful or unsuccessful).

Not tested = I haven't try this configuration even if it possibly could have been done.

NA = Not applicable: (I personally believe that) this is not a possible (directly operable) configuration.
Partial success only: random failure observed
Systematic failure case demonstrated
Systematic Success

Notes:

(*)When Control Points are assigned and when alignment optimization is done manually PTGui and PanoTools give a very excellent result as long as the lens has been calibrated with more than two images prior to attempt in stitching DS this way. I could get the best after calibration with for images shot 90 degrees apart.

(**) In fact I have imported the script file from PTGui here to dodge a second optimization and to test the very same warped panorama for blending.

(***) This line could be removed when the problem reported in Part 3 shall be resolved as it only examplify the yet unexplained phenomenon.

Highlights

1) From this table, one can also conclude that from all these programs, only Panoweaver seems to be able to consistently auto-stitch Double Shot panoramas. Remember: this is a dedicated program that was designed and optimized for this. StitcherU-DS seems not up to deliver what it was intended for though.

2) Automatic three shots stitching is viable with most of the tested software but AutopanoPro looks like unable to perform consistently well with such a type of fisheye circular images when the number of those is less than four. AutopanoPro has only recently been updated to handle fisheye images.

3) Stitching four images shot 90 degrees apart doesn't give any problem with any of the programs at least with the set of sample images that I was kindly delivered by Roberto Gomez Torres

Conclusions (Part 2)

Panoweaver 5.0 associated with the Smartblend plugin seems to be the only program that fits the requirements of true DS panography. All others that were tried failed in a way or an other and especially they could not make automatic CP detection and images alignment reliably and consistently. Panoweaver 5.0 is unfortunately only available for Windows (Panoweaver 4.0 for MacOS is not supported nor maintained anymore. Additionnally it is awfully slow and doesn't accept blender plugins).

It is however possible to overcome this hurdle and one could output DS panoramas as good as with Panoweaver with other programs, but this require some skill, experience or high end technical support. Lens calibration and use of templates may even help to get outputs as fast as with the EasyPano's software. The problem is that the targetted professionals that are supposed to use DS panography are generally not the more technically oriented people.

In addition, from Part 1 it has been shown that with the source images from the Sunex Superfisheye or the Sigma 4.5-mm, the quality of the DS panorama (whatever the way to make it happen) may be judged honestly rather poor along the sole seam of the stitch. This is still true when the Sigma lens apperture is closed down to it sweet spot at around f/11. I would therefore not recommand to use those lenses for DS panography if more than three source images can be shot instead of two. Additionnally this opens dramatically the number of possible stitching programs!
But then another question arises: what is the benefit brought by such very short focal fisheye lenses compared with other means? Why not use a Sigma 8-mm f3.5 on the APS-C DSLR despite the FOV strictly limited to 180 degrees and take four shots? Why not use a (shaved) Tokina 10-17-mm or a Nikkor 10.5-mm on a FF DSLR camera and take three shots?

The latter combination(s) allow to shoot as few as three photos for a full spherical panorama with much more resolution and sharpness than the same withthose two very short focal lenses.

There may be other reasons to prefer using these exotic lenses but they have to be very personnal and cannot be a general recommandation IMHO.
This study has convinced me that the IPIX era is definitly over and that attempting to stitch DS panorama is not really viable anymore. Better quality shall be got easily and as fast with at least three source images and the modern means that are nowdays affordable just put the DS idea to become a concept of the past.

Michel Thoby

17 February 2008