Smearing top of 360 pano's. [Archive] - Nodal Ninja Forum

PDA

View Full Version : Smearing top of 360 pano's.



Ben Rubinstein
06-04-2011, 12:08 PM
Hi,

Using the NNP points from the website I'm getting a weird smearing at the top of files shot with the Sigma 8mm f3.5 fisheye. This looks like a circle with spikes radiating from it in a virtual tour. I can't seem to get rid of it. I get this with my 16-35L as well though less.

Any suggestions?

Many thanks!

Ben Rubinstein
06-04-2011, 01:44 PM
The first pano here is a good example:

http://www.studio-beni.net/Town.html

Have a look at the zenith. This was shot with 4 shots at 0 degrees then one zenith shot but it's the same with 3 shots with/without a zenith shot.

hindenhaag
06-04-2011, 09:32 PM
Hi Ben,

I feel you are clearly out of NPP. The camera lens axis is moved around the center point of the rotator. When you move down to the nadir, you clearly see a hacksaw cutting to the right. This means you have to move the camera body to the right till you get a round tripod head.


193






http://www.easypano.com/forum/display_topic_threads.asp?ForumID=1&TopicID=4162

http://www.johnhpanos.com/epcalib.htm

Besides following John's comment, I would recommend to try these links to recheck your NPP for lower and upper rail setting. Fortunately there sometimes are faults in the lower and upper rail settings in the Database. What camera do you use?

Regards,
Heinz

BTW: in this tripod postion to the sun you get shadows of 3 legs. When you turn it around a little bit, you could match the shadow of two legs to one, which gives less work to get rid of the shadow.

John Houghton
06-04-2011, 09:34 PM
This is a blending problem. Most blenders don't support blending the zenith and nadir areas. PTGui's own blender usually gives good results at the zenith but Smartblend tends to give results as in your example. Solutions:

1. Take a zenith shot and make sure it gets used by contriving an actual hole at the zenith left by the other images.
2. Do two stitches if necessary - one with PTGui's blender and one with Smartblend and merge the two.
3. Rotate the panorama down by 90 degrees to bring the zenith (and nadir) to pitch=0 where they will get properly blended. Then remap the output image back to its normal orientation.

John

Ben Rubinstein
08-20-2011, 01:32 PM
Hi Ben,

I feel you are clearly out of NPP. The camera lens axis is moved around the center point of the rotator. When you move down to the nadir, you clearly see a hacksaw cutting to the right. This means you have to move the camera body to the right till you get a round tripod head.


193






http://www.easypano.com/forum/display_topic_threads.asp?ForumID=1&TopicID=4162

http://www.johnhpanos.com/epcalib.htm

Besides following John's comment, I would recommend to try these links to recheck your NPP for lower and upper rail setting. Fortunately there sometimes are faults in the lower and upper rail settings in the Database. What camera do you use?

Regards,
Heinz

BTW: in this tripod postion to the sun you get shadows of 3 legs. When you turn it around a little bit, you could match the shadow of two legs to one, which gives less work to get rid of the shadow.

Hi,

I moved the lower rail to the right by a full 1.5cm at 1mm intervals, no difference. :(

Ben Rubinstein
08-20-2011, 01:36 PM
This is a blending problem. Most blenders don't support blending the zenith and nadir areas. PTGui's own blender usually gives good results at the zenith but Smartblend tends to give results as in your example. Solutions:

1. Take a zenith shot and make sure it gets used by contriving an actual hole at the zenith left by the other images.
2. Do two stitches if necessary - one with PTGui's blender and one with Smartblend and merge the two.
3. Rotate the panorama down by 90 degrees to bring the zenith (and nadir) to pitch=0 where they will get properly blended. Then remap the output image back to its normal orientation.

John

I don't have the smartblend plugin, just been using the regular PTGUI one though I get exactly the same results with APP Giga if that helps.

I just tried your method, I cut the 3 regular (0 degrees) pano's so that the program was forced to use the zenith, etc shot, no luck, still got the smearing.

EDIT Just downloaded Smartblend, looks exactly the same.

Ben Rubinstein
08-20-2011, 02:37 PM
OK, just out of interest I tried doing the pano with a simple ball head rotating around the base, i.e. without the nodal slides. The nodal errors were awful but aside from that, the smearing was exactly the same!

Something very weird going on...

hindenhaag
08-20-2011, 03:17 PM
Hi Ben,

Can you send the set of pics?

Heinz

Ben Rubinstein
08-21-2011, 12:57 AM
The files are here, www.studio-beni.net/test.zip

Camera used is a Canon 5D with Sigma 8mm f3.5 fisheye. The correct position for the lower rail with this setup is 5.3mm according to the NN figures and my tests, I'm using an RRS L plate with the NN Arca adaptor, both of which together add exactly a 9mm offset to give a correct position of 6.2mm (Please note that using the NN5 with the CP-1 at 53mm gives the same smearing!) The test images run from 5.5mm to 7.5mm with an additional 4mm, 4.5mm and 5mm included, as well as the same pano shot without the NN5 setup. Position of the camera was +5 degrees tilt to ensure that the ceiling was captured in full without the need for a zenith shot (not that a zenith shot makes any difference, I tried that often enough). Three frames were used in each case.

The pano's were made in PTGUI with the PTGUI stitching engine apart from one of the images which I used the Smartblend engine for and is labled accordingly. None of the images are in any way sharpened.

Heinz, I really appreciate your time and effort to look at these pictures!

hindenhaag
08-21-2011, 08:38 AM
Hi Ben,

I just checked the set of your pics. First thing I'd like to say you have to check your lens in front and the rear for dust or spots on it. The dark dots in the panos are spots on the lens or on your sensor. Please try to clean both of them.

What do you means by smearing? I suppose the dots on the ceiling as well as dots on the floor in some of the panos. In some cases you get stitching errors near your tripod, or on the upper wall.

Do you have a set of LRS pics taken according to Smooth method? Plus a set of pics to check your upper rail setting without sending a pano? With a reference point checked against a vertical line? Like explained in the links we send you. This is the basic od a set up which should be right.

To check what happened we have to get the set of pics taken for a single panorama to stitch it ourselves when we should try to help you. So please try to send these set of pics. Then I can check what is going wrong.

Regards,
Heinz

Ben Rubinstein
08-21-2011, 08:54 AM
Hi Ben,

I just checked the set of your pics. First thing I'd like to say you have to check your lens in front and the rear for dust or spots on it. The dark dots in the panos are spots on the lens or on your sensor. Please try to clean both of them.

What do you means by smearing? I suppose the dots on the ceiling as well as dots on the floor in some of the panos. In some cases you get stitching errors near your tripod, or on the upper wall.

Do you have a set of LRS pics taken according to Smooth method? Plus a set of pics to check your upper rail setting without sending a pano? With a reference point checked against a vertical line? Like explained in the links we send you. This is the basic od a set up which should be right.

To check what happened we have to get the set of pics taken for a single panorama to stitch it ourselves when we should try to help you. So please try to send these set of pics. Then I can check what is going wrong.

Regards,
Heinz

Hi,

Yes the sensor is dirty, sorry.

Have a look at the top of the frame, accentuate contrast and sharpness and you'll see the smearing, it's not so apparent on a white wall but with blue sky or a non white surface it's horribly ugly.

Ben Rubinstein
08-21-2011, 09:00 AM
I will try the smooth method now, thanks!

hindenhaag
08-21-2011, 09:34 AM
Ben,

you do not have to say sorry. You ask a question and we, the forum, tries to respond. This way everyone of us has learned step by step. Asking, getting hints what might be wrong or getting answers with definite solutions.

Sorry to say, my D700 was stolen and I just lend one from NPS = Nikon Professional Services to find out settings for the new M1-Line. For fisheye lenses, Nikkor 14-24, 24-70, 70-200, 50. I will change it tomorrow for a D3X.

I will have a look to your pics concerning contrast and sharpness the next days. Ok?

Cheers,
Heinz

Ben Rubinstein
08-21-2011, 09:42 AM
Ouch! I hope you were insured but that is one heck of an upgrade to what it without a doubt the most superior 35mm DSLR in existence today! I hope your computer is up to stitching those big georgous files...

I'm midway using the 'smooth' method, so far I've managed to accomplish a perfect circle at the base and am trying to work out if the upper arm setting is accurate. I had made the mistake of believing the NN settings, infact it was the reason I chose NN so that I wouldn't have to go through all this testing (more fool me for trying to get out of it!). Once I have the settings perfect I will see if the smearing still exists!

hindenhaag
08-21-2011, 10:02 AM
Ben,

Take your shots around with -15 pitch. Import to PTGui Pro. Align, Advanced top right, Optimizer, if you do not use the latest Beta "minimize lens distortion" set this to "Heavy + lens shift". Run Optimizer. Click "OK" for the result with the opened window.

Go do the tab list on the very top : "Control Points". Choose for "Delete Worst Control Points". "Optimze" again and go back to "Project Assistant". In the opened window choose "Create Panorama". Before that check your panorama settings box: Projection has to be set to "Equirectanguar"."Field of View to 360 (hor) and 180 (vert) . Otherwise you do not see the whole tripod. Set "file format" to "Quick Time VR" and then Create Panorama. Go to "View" to see the result. For the best you have Quick Time 7 installed on your computer to see the result. Once opened, move down to the nadir to see your rotator totally and zoom in to check for Teeth on the circle saw. I f the tooth cuts to the right, move the camera to the right, if it cuts to the left, move to the left. Try to find the position without a tooth and check from right to left where then the cutting direction of the teeth should change from right to left. At least you might have to check for 0,5 mm from right to left.

Success,
Heinz

Ben Rubinstein
08-21-2011, 10:06 AM
Hi,

I had been doing all of that, though according to the 'smooth' article with the sigma I don't need any minus pitch. I've been optimizing manually actually and creating the pano then viewing within Panotours. I seem to have found 6.5mm as my perfect setting without any teeth, the 'circle' is still somewhat angular however so I'm working along the top rail trying to find the correct setting there to get a real circle.

hindenhaag
08-21-2011, 10:22 AM
This is how it should look like:

260261262

Success,
Heinz

hindenhaag
08-21-2011, 10:34 AM
Ups, another set which is complete from cut to right , ok, cut to the left:

263264265

You have to look close to see the very small cutting teeth, specially on the last one.

Heinz

Ben Rubinstein
08-21-2011, 10:57 AM
OK I think I've cracked it, here are the RAW's www.studio-beni.net/RAW.zip sorry for the lower resolution, it's only a 5D, I can't see any teeth and this is as circular as my testing can get me so far. If I should be testing to within half mm levels then let me know please. I'm still getting the smearing at the top of the frame unfortunately, the resolution near the edges of the fisheye becomes very very soft and I'm wondering if that is what is causing it. I've tried using the above NPP points but shooting -15 degrees around with a zenith shot and see if that helps.

Wim.Koornneef
08-21-2011, 11:59 AM
Hello Ben,

Is it OK if I chime in ?

I have experience with the 5D and Sigma 8 lens and I can confirm that the lens is not as sharp at the edges as you wish, the result is a soft nadir and zenith when you only shoot roundshots.

A zenith and nadir shot can enhance the result but you can also can get a much better result when you sharpen the outerpart of the images a bit.
You can do this with Photoshop, use a circular selection tool with a large feather of approx. 50 px and select a circle of approx. 3/4 of the diameter of image circle, inverse the selection and sharpen the image.

BTW, 6 full circular images is a bit of overkill with this camera and lens combination, even with 3 images around it is possible to stitch a fine panorama and for a good overlap, a sharp image and easy stitching 4 around is really sufficient.
When you shoot 4 around (instead of 6 around) you have to move the camera approx. 2-3 mm backwards to get the lens in the proper NPP position.

Tip: when using only roundshots a tilt of approx. +5 degree will give you a much better coverage in zenith.

Success,
Wim

Ben Rubinstein
08-21-2011, 12:14 PM
Hi Wim,

The smearing problem I'm having is even when I'm using a zenith shot so that the periphery of the fisheye circle is not being used, it's extremely fustrating.

Here is a crop to show the problem (heavily accentuated as it's not as apparent on a white surface):
http://www.studio-beni.net/zenith-problem.jpg

This is 6 frames at -15 degrees plus a zenith though the problem is identical with 3 frames at +5 degrees and no zenith shot.

Wim.Koornneef
08-21-2011, 12:21 PM
Hello Ben,

Is the crop a part of the equirectangular ?

If so this is not smearing but just the way the pixels in zenith (and nadir) are normally stretched.
You should only judge the quality of the nadir and zenith of the pano when the equirectangular is remapped to cubes or viewed as Flash or QTVR pano.

BTW, imo your image looks a bit oversharped (perhaps this is due the accentuating you applied).

Wim

Ben Rubinstein
08-21-2011, 12:25 PM
Not a bit, hugely! :-) I had to oversharpen and over contrast the image to show the pattern in what is just a white blank surface.

If you look at the link at the very beginning of this thread the effect of this smearing is having at the zenith with this spiky 'bellybutton' effect.

Wim.Koornneef
08-21-2011, 12:31 PM
Hello Ben,

The first link in the thread gives a 404 error...
(http://www.studio-beni.net/Town.html)

You didn't answer my question of the crop was taken out of a equirectangiular, this is important to know as I suspect that if this is part of the equi nothing is wrong...

Wim

Ben Rubinstein
08-21-2011, 12:34 PM
It's a 360X180 if that helps and it's the top 10% of the frame. Sorry not sure what you were getting at?

Wim.Koornneef
08-21-2011, 12:50 PM
Hello Ben,

So the crop is from a 360x180 --> equirectangular --> stretched pixels by nature in zenith and nadir--> always a smearing effect when viewing an equirectangular at nadir and zenith.

You really should remap the equirectangular to a cubic pano to judge the quality of nadir and zenith.
You can do this with PTGui by converting the equirectangular to Flash or HtmlCss3 pano or by outputting the pano as QTVR instead of equirectangular.

I have stitched your example images and the output looks fine to me (except for a small hole in zenith and the many bad spots caused by dirt on sensor and/or lens) so I think you are a bit to critical about the quality...
The NPP setting looks fine to me, very low CP errors and no stitching errors so you got it right.

Wim

John Houghton
08-21-2011, 03:38 PM
Ben, Is this shot of the zenith showing what you call "smearing" : http://tinyurl.com/3otejzz . I think the radial streaking is the stretching of noise due to the image warping at the edge of the image circle. The sharpening you have applied has accentuated the effect. If you apply some noise reduction and don't sharpen in ACR, then the smearing will be considerably reduced. Shooting at a lower ISO setting (with longer exposures) will reduce the noise too.

John

Ben Rubinstein
08-21-2011, 11:42 PM
Ben, Is this shot of the zenith showing what you call "smearing" : http://tinyurl.com/3otejzz . I think the radial streaking is the stretching of noise due to the image warping at the edge of the image circle. The sharpening you have applied has accentuated the effect. If you apply some noise reduction and don't sharpen in ACR, then the smearing will be considerably reduced. Shooting at a lower ISO setting (with longer exposures) will reduce the noise too.

John

I'm getting it even when the zenith shot is the only one used for that part of the zenith, i.e. it's the center of the frame rather than the edge. I'm going to do a test outdoors in a couple of hours (bit tied up now) and post up the results with what should be mm perfect NPP according to the smooth method. Then you guys can tell me what you think.

Wim.Koornneef
08-22-2011, 12:12 AM
Hello Ben,

It could be very well that over sharpening in ACR is the reason of the pattern John showed in his screenshot.
I didn't see this pattern in my output because I processed the RAWs with Canon's DPP software with a moderate sharpening (using a non canon lens with DPP is only possible after modifying the Exif info to fool DPP).
The pattern itself is normal for many other fisheye lenses when using only roundshots but is not disturbing if the noise is removed and the image is not oversharped.

The grey spot in the center of John's zenith image is caused by the dark edge of the circular fisheye image, with a smaller crop circle the dark spot is gone but then you have a small hole in zenith, that is the reason why I suggest to use a tilt of +5 degree to get a better coverage of zenith.

You said that the smearing also occurred in a pano when using rounshots with a downtilt of 15 degree and a zenith shot, the radial pattern John showed is then not possible as there is no overlap of the roundshots in zenith so I am puzzled what the smearing is you are talking about.
If you show a screenshot, or a crop, of the cubic pano then perhaps that can make clear what the smearing is.

Wim

Ben Rubinstein
08-22-2011, 02:16 AM
Right,

Here is an example of the problem I was having with the smearing showing up in regular pano's at the zenith (not on first pano but pretty much all the others) http://www.studio-beni.net/Bart.html This was using the NPP values provided by NN.

Now here is a VT showing two pano's, one shot with zenith and one without, now using the NPP's obtained with the 'smooth' method that Heinz had recommended. The smearing is still there but not showing up with regular processing and the zenith now looks clean: http://www.studio-beni.net/New.html

I've just gone and shot an HDR of the same scene to see whether the HDR processing is accentuating the smearing. HDR is very important to my shooting as most of the times I get into properties it's the middle of the day and I need to preserve window view details. I'll report back as soon as I've done them!

John Houghton
08-22-2011, 02:40 AM
Ben, The pointy star at the zenith that you are calling "smearing" is often referred to as a vortex and is a blending problem. See my early response dated 06-05-2011 05:34 AM. It has nothing directly to do with the NPP setting. Are you using PTGui and its own blender?

John

Wim.Koornneef
08-22-2011, 02:43 AM
Hello Ben,

The example panos are fine to me, you say there is smearing in it but not visible with regular processing so I think the smearing problem (whatever is was) is tackled.
I think removing of the dust on lens and/or sensor and removing/reducing the remaining CA is much more important to gain image quality...

Success,
Wim

PS. John, are you sure you see a vortex in Ben's 2 examples, or are you referring to a different pano ?
http://www.studio-beni.net/New.html

Ben Rubinstein
08-22-2011, 02:45 AM
Ben, The pointy star at the zenith that you are calling "smearing" is often referred to as a vortex and is a blending problem. See my early response dated 06-05-2011 05:34 AM. It has nothing directly to do with the NPP setting. Are you using PTGui and its own blender?

John


I was John, only downloaded the other engine for the first time yesterday.

Ben Rubinstein
08-22-2011, 02:46 AM
Hello Ben,

The example panos are fine to me, you say there is smearing in it but not visible with regular processing so I think the smearing problem (whatever is was) is tackled.
I think removing of the dust on lens and/or sensor and removing/reducing the remaining CA is much more important to gain image quality...

Success,
Wim

Thanks Wim, the dust/CA will of course be gone on commercial output, I just didn't bother for this testing.

hindenhaag
08-22-2011, 03:29 AM
BTW: Thx Wim and John for stepping in. :001_smile:

You got two of the very experts of this forum to get rid of your problem Ben.

Success,
Heinz

John Houghton
08-22-2011, 03:32 AM
PS. John, are you sure you see a vortex in Ben's 2 examples, or are you referring to a different pano ?

Wim, I'm referring to the several panos (except the first) at http://www.studio-beni.net/Bart.html .

John

Ben Rubinstein
08-22-2011, 03:37 AM
Well! Seems the problem was the HDR program! The weird twisty thing is back again in all its glory! Trying to work out what it's doing that is causing that but it's in line with John's theory that the vortex is software not hardware based. The HDR program is somehow warping the finished stitch. I'm going to try HDR-ing the individual frames pre-stitch to see if that solves the problem.

Oh and I hadn't forgotten my own thanks for all of you who have been so patient with me! I so wanted to believe in my NN5 and thanks to you I now can!

John Houghton
08-22-2011, 03:40 AM
Ben, I'm a little surprised to hear that these vortices are being produced with PTGui's blender, though possibly they are somehow due to subsequent processing. If you care to upload a set of camera images that illustrates the problem (one exposure set only, half size jpegs will do), maybe we can get a better idea of what is going on and be in a better position to offer a solution.

John

-Edit-
I gather from the last response of Ben that my guess was right and HDR is the culprit.

Ben Rubinstein
08-22-2011, 03:48 AM
The HDR program seems to be changing the ratio though only by a pixel or so and that's causing the vortex. I tried filling in the pixel gap with black at the bottom of the frame but it doesn't seem to help though I don't quite understand why.

Ben Rubinstein
08-22-2011, 04:34 AM
OK, here is an HDR'd version using the same HDR software but doing the HDR at the single image stage rather than with the stitched file. Hey presto it's fine!

www.studio-beni.net/New2.html

Thanks again for all the expertise and most of all for the patience especially as I was barking up the wrong tree for 95% of the time! I really appreciate your help!