Suggestion

[Michael Havens]

Hello gentlemen (and women),

I take pictures of the interior of houses. I use hugin in lew of purchasing a wide angle lens (I do not have a spare $1000 dollars laying around). I can only offer an idea, I don't have the technical aptitude for anything else (wish I did). Now it is good but you can see the seams if you look for them. What maybe someone could do is make it so control points from images proportionally scale to  fit together. Or perhaps I am doing something wrong.

Here are the photos and stitched picture.

https://drive.google.com/open?id=0B2xvsVTZy4y1T0NsMkdOdGo2TWs

[Sean Greenslade]

On August 7, 2016 12:02:16 PM EDT, Michael Havens <havens....@gmail.com> wrote:
>Here are the photos and stitced picture.
>https://drive.google.com/open?id=0B2xvsVTZy4y1T0NsMkdOdGo2TWs

>
>On Sunday, August 7, 2016 at 11:32:43 AM UTC-4, Michael Havens wrote:
>>
>> Hello gentlemen (and women),
>>
>> I take pictures of the interior of houses. I use hugin in lew of
>> purchasing a wide angle lens (I do not have a spare $1000 dollars
>laying
>> around). I can only offer an idea, I don't have the technical
>aptitude for
>> anything else (wish I did). Now it is good but you can see the seams
>if you
>> look for them. What maybe someone could do is make it so control
>points
>> from images proportionally scale to fit together. Or perhaps I am
>doing
>> something wrong.
>>

There are many possibilities. If you took the photos handheld, or used a tripod without a nodal panning adapter, there could be uncorrectable parallax in the photos. Since there are objects in the near field, this is an important point to consider.

Have you also gone through the entire optimization process? After finding control points, verifying that they are mostly all correct, then doing all the various optimization steps, including lens distortion. What was the final RMS error of the last optimizer run?

--Sean

[Stefan Peter]

Hi Michael

On 07.08.2016 18:02, Michael Havens wrote:
> Here are the photos and stitced picture.
> https://drive.google.com/open?id=0B2xvsVTZy4y1T0NsMkdOdGo2TWs

From what I understand from the pics you have posted (and thank you for
not sending them out to everyone on this mailing list but providing a
link for downloading), I think your main problem is exposure or dynamic
range of a picture. Your leftmost image features a very bright window
making up an estimated 40 percent of the image and the interior wall
making up the rest. You camera will try to level out these two extremes
so that the average will be 19% grey, resulting in the window to be way
overblown and the wall to light, too, if compared with the middle shot
that has now window. Remember that your eyes can cover a dynamic range
of up to 18 steps but your digital camera can not.

I'd recommend the following:

o Learn how to shoot the left shot so it looks natural. You will
have to use a technique known as bracketing. This means shooting
manually and doing -6, -4, -2, 0, +2, +4, +6 exposed images of
the window. Then, you will have to combine the shoots using
either hugin or enfuse or some other hdr tool to get an image
with reduced dynamic range. If the -6 to +6 range does not
lead to satisfactory results, feel free to extend the range or
widen the steps between exposures.
o Another issue may be the light color differences between the
outside (sunlight) and the inside (tungsten?) of the room.
You may have to prepare for these differences when preparing
the images for enfuse or your HDR tool of choice.
In the past, I have been known for using two versions of the
same image, one developed for the outside color and masked
to only include these areas and a copy thereof developed for
the inside lighting, excluding the rest of the image. In such cases,
Hugins masks feature comes handy!
o Always shoot in RAW mode if your camera supports it. This allows you
to take advantage of the full dynamic range your camera supports and
selectively apply colour profiles depending of the location of the
best exposed part of the image you are handling.

Please do not refrain to come back to us if you have additional
questions or want constructive feedback concerning your work.

With kind regards

Stefan Peter


--
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
(See https://en.wikipedia.org/wiki/Posting_style for details)

[Michael Havens]

I found the control points. Of the three pictures one of them would not do it automatically so I inserted a couple manually. As for verifying the ones it found automatically I didn't inspect them other than glancing at them and seeing they appeared correct.

How do I figure out lens distortion? Darktable knows my lens including it's distortion. Can I use that information somehow?

[Michael Havens]

I figured out lens distortion but what should I take a picture of? I think that image with lines radiating from a central point. What do you know? I just took a picture of my monitor. with that it found two x axis lines and one y.

[Terry Duell]

On Mon, 08 Aug 2016 00:48:24 +1000, Michael Havens
<havens....@gmail.com> wrote:

> Hello gentlemen (and women),
>
> I take pictures of the interior of houses. I use hugin in lew of
> purchasing
> a wide angle lens (I do not have a spare $1000 dollars laying around). I
> can only offer an idea, I don't have the technical aptitude for anything
> else (wish I did). Now it is good but you can see the seams if you look
> for
> them. What maybe someone could do is make it so control points from
> images
> proportionally scale to fit together. Or perhaps I am doing something
> wrong.
>

I had a look at the stitch you put on Google drive.
I can see the effect of alignment errors, and these are probably due to
parallax errors.
You may be able to improve on this result by editing your control points
and possibly manually adding some control points.
I'll take a look at what I can do with your images and provide further
advice if I can.
It is always helpful to provide your project file (.pto) when asking for
help as the project file can provide a lot of useful info.

Cheers,
--
Regards,
Terry Duell

[Terry Duell]

On Mon, 08 Aug 2016 08:52:21 +1000, Terry Duell <tdu...@iinet.net.au>
wrote:

> I'll take a look at what I can do with your images and provide further
> advice if I can.

Not much help with that.
The images I got off Google drive were of such poor resolution that
finding control points either automatically or manually is very
problematic.
I see others have provided advice.

[Sean Greenslade]

On August 7, 2016 6:23:04 PM EDT, Michael Havens <havens....@gmail.com> wrote:
>I figured out lens distortion but what should I take a picture of? I
>think
>that image with lines radiating from a central point. What do you know?
>I
>just took a picture of my monitor. with that it found two x axis lines
>and
>one y.

One of the nice things about Hugin is its ability to back-calculate lens parameters. If you place a lot of good, varied control points in the overlap regions, then enable the lens distortion optimizations, Hugin should be able to find and correct the distortions of your lens. This works best with a tripod & nodal rotation, however. Parallax errors in the source photos can cause the distortion parameters to go crazy during the optimization.

--Sean

[Greg 'groggy' Lehey]

On Sunday, 7 August 2016 at 7:48:24 -0700, Michael Havens wrote:
On Sunday, 7 August 2016 at 9:04:11 -0700, Michael Havens wrote:

Once is enough.

> I take pictures of the interior of houses. I use hugin in lew of purchasing
> a wide angle lens (I do not have a spare $1000 dollars laying around). I
> can only offer an idea, I don't have the technical aptitude for anything
> else (wish I did). Now it is good but you can see the seams if you look for
> them.

You can certainly do a lot better than that.

> What maybe someone could do is make it so control points from images
> proportionally scale to fit together.

Others have addressed your probable issues, but I'm still left
wondering what you mean with this suggestion. Can you elaborate?

On Sunday, 7 August 2016 at 17:47:42 -0400, Michael Havens wrote:
> I found the control points. Of the three pictures one of them would not do
> it automatically so I inserted a couple manually. As for verifying the ones
> it found automatically I didn't inspect them other than glancing at them
> and seeing they appeared correct.

The big issue in these images is exposure, both overall and in
relation to each other. This is part of the reason why no control
points are detected between the left-hand two images. Another reason
could be the relative lack of appropriate features for control points.
It's not a good idea to have the camera on auto ISO, but it's unlikely
that this could cause significant problems with control point
detection.

> How do I figure out lens distortion? Darktable knows my lens
> including it's distortion. Can I use that information somehow?

Your camera has a raw format (NEF), and that's what you should be
using. You can probably do this automatically with Nikon ViewNX-i,
available for free at
http://downloadcenter.nikonimglib.com/en/download/sw/71.html. But I
greatly doubt that distortion is your problem. If you want to use
Darktable, process the image to correct distortion and feed the output
images to Hugin.

Since this doesn't look like a once in a lifetime, never to be
repeated shot, and you haven't mentioned entrance pupils, I'd suggest:

- Use a tripod.
- If you have a panorama bracket, use it. Otherwise mount the camera
on a rail. It seems that you can get entrance pupil information for
your setup from
http://hugin.sourceforge.net/docs/manual/Entrance_pupil.html and
http://wiki.panotools.org/Entrance_Pupil_Database.
- Use manual focus, manual exposure (check which is most appropriate
before you start), recommended ISO setting for your camera, and raw
images.

Greg
--
Sent from my desktop computer.
Finger gr...@FreeBSD.org for PGP public key.
See complete headers for address and phone numbers.
This message is digitally signed. If your Microsoft mail program
reports problems, please read http://lemis.com/broken-MUA

[Michael Havens]

 thank you all for your assistance. I like the idea of using dt to correct distortion.  but I just spent a little time playing with the lens calibration. How should I reset that back to its original state?

While I mentioned that I tried calibrating this with a starburst image. The corrected image was a fisheye. Why did I get that result?

[Michael Havens]

> What maybe someone could do is make it so control points from images 
> proportionally scale to fit together. 

Others have addressed your probable issues, but I'm still left 
wondering what you mean with this suggestion.  Can you elaborate? 

Now that I think of it I don't know what I was thinking! What I meant by that is you have control point 'x' and 'y' on one picture and you have the corresponding 'x' and 'y' points on the other picture. but the second x/y points were taken at a different zoom factor... you know what I mean now?

[Sean Greenslade]

>> Now that I think of it I don't know what I was thinking! What I meant
>by
>that is you have control point 'x' and 'y' on one picture and you have
>the
>corresponding 'x' and 'y' points on the other picture. but the second
>x/y
>points were taken at a different zoom factor... you know what I mean
>now?

That's fine, Hugin can handle that. Two things, though: one, make sure the photos have different lenses assigned to them, that way the optimizer knows their FOVs can be different. And two, know that the control point finder may get confused by that, so you may have to do more manual control point adding / correction.

--Sean

[Frederic Da Vitoria]

I am not sure you realize the importance of parallax errors. By visually comparing 222 with 223, I notice the chair does not cross the same part of the floor in the 2 pictures. I don't think any panorama software will be able to correct this. Same issue between 223 and 224, the intersection between the wheel and the seat behind it. You really need to take pictures so that those errors are as small as possible. The best way, of course, is to use a suitable tripod. At least, even if you don't or can't buy a tripod, you have to take care to turn your camera around itself instead of turning it around you.

The second point to consider is the exposure issue. 222 was taken with completely different parameters (ISO 280) from the others (ISO 1800 and 2200). (BTW, I don't understand the parameters: 1/30, F3.5, I don't understand why the camera needs to go to ISO 280, I don't understand ISO 2200 more). In my experience, Hugin does not do a very good job about correcting different exposures like this. I suggest you should work on 222 to correct the exposure before trying to assemble. If you shot in RAW, you should be able to achieve a good correction.

--

Frederic Da Vitoria
(davitof)

Membre de l'April - « promouvoir et défendre le logiciel libre » - http://www.april.org