How much overlap for TuFuse to invoke focus blend?

Discussion forum for Tawbaware's TuFuse and TuFuse Pro software
Post Reply
Gr8Scot2
Posts: 8
Joined: Fri Sep 04, 2009 9:47 pm
Location: Washington State

How much overlap for TuFuse to invoke focus blend?

Post by Gr8Scot2 » Tue Nov 03, 2009 5:34 am

Does anyone now how much overlap must occur between 2 photos before Tufuse will invoke focus blending?

I guess what I am asking is if, for example, I take a three photo wide panorama, then I go back and retake three more of the same scene but with a different focus setting, would Tufuse perform focus blending across the whole panorama (assuming PTassembler's blending options were set correctly)?

In a case like this, any one of the first three photos would not exactly line up with any of the second set of differently focused photos. Maybe they will line up 75% or maybe only 60% if I am really sloppy. Would they have to line up within like 90% or something for focus blending to work? Is there a threshold or can Tufuse do focus blending on only a small slice of 2 photos that overlap?

I hope this question makes sense. It is kind of difficult to ask in only a few words with no diagrams.

Thanks in advance for any help!
-Scot :?
-Scot Thomas

Bart
Posts: 44
Joined: Sat Jun 10, 2006 8:59 am

Re: How much overlap for TuFuse to invoke focus blend?

Post by Bart » Tue Nov 03, 2009 6:21 am

Gr8Scot2 wrote:Does anyone now how much overlap must occur between 2 photos before Tufuse will invoke focus blending?

I guess what I am asking is if, for example, I take a three photo wide panorama, then I go back and retake three more of the same scene but with a different focus setting, would Tufuse perform focus blending across the whole panorama (assuming PTassembler's blending options were set correctly)?
Hi Scot,

This is no problem, as long as all of the images are taken from the same view/perspective point (which may be hard if the exact position and height parameters weren't marked).

TuFuse doesn't mind what's in the images, it just compares the level of detail between images at a given corresponding pixel position. At relatively close distances that also means that the differences in magnification caused by different focus settings need to be compensated for, e.g. by optimizing for the "focal length" per image.

When shooting the images it's best (most natural looking) to make sure that the DOF for each shot overlaps a little between shots for focus blending. Depending on the shooting scenario, that can be calculated in advance into a number of focus distances at a given aperture. The difficulty is sometimes how to determine those distances when focusing (I use a laser range finder for that), although cameras capable of "LiveView" have an advantage when it's done visually.

Cheers,
Bart

maxlyons
Posts: 3649
Joined: Fri Jun 20, 2003 8:55 pm
Location: USA
Contact:

Re: How much overlap for TuFuse to invoke focus blend?

Post by maxlyons » Tue Nov 03, 2009 9:55 am

Gr8Scot2 wrote:Does anyone now how much overlap must occur between 2 photos before Tufuse will invoke focus blending?
If you don't tell TuFuse explicitly what type of blending you'd like it to do, then it tries to guess the appropriate type of blending by looking at the relative brightness of overlapping pixels. By default, if the images differ by one third of a stop, TuFuse assumes that you want to do exposure blending. If they differ by less than one third of a stop, then it assumes you want to do focus blending. (you can configure the threshold using the "-e" switch). The relative positions of the images (i.e. how much they overlap) isn't relevant to this decision.
I guess what I am asking is if, for example, I take a three photo wide panorama, then I go back and retake three more of the same scene but with a different focus setting, would Tufuse perform focus blending across the whole panorama (assuming PTassembler's blending options were set correctly)?
I guess we should be clear what we are talking about:

1. TuFuse as invoked from the command line.
2. TuFuse as invoked via PTAssembler.
3. TuFuse Pro.

It sounds like you are talking about #2.

In that case, the relative position of the images does matter in the sense that PTAssembler invokes TuFuse for "stacks" of images before blending those fused stacks into a final composite. It does not do the opposite (i.e. create three panoramas, and then fuse those three panoramas into a final composite). And, PTAssembler will invoke TuFuse for images that you have identified as belonging to the same stack (see step 1 of PTAssembler).

Max

Gr8Scot
Posts: 5
Joined: Wed Apr 02, 2008 2:45 pm

Post by Gr8Scot » Tue Nov 03, 2009 8:35 pm

Ok, so if I have 2 images that are overlapped by 50% and each have a different focus and if I have set up PTassembler to do only focus blending and to auto detect stacks, will it perform any focus blending?

If not, then how much more overlap would be needed before it would perform focus blending? Or should I manually declare both pics as a "stack" even though the overlap is only 50%?

-Scot

maxlyons
Posts: 3649
Joined: Fri Jun 20, 2003 8:55 pm
Location: USA
Contact:

Post by maxlyons » Wed Nov 04, 2009 1:01 am

Gr8Scot wrote:Ok, so if I have 2 images that are overlapped by 50% and each have a different focus and if I have set up PTassembler to do only focus blending and to auto detect stacks, will it perform any focus blending?
No...they are too far apart for PTAssembler to guess that they belong to the same stack.
If not, then how much more overlap would be needed before it would perform focus blending?
A lot. The logic that PTAssembler uses is as follows:

diff_threshold = (0.05 / 2) * (image1.lens_FOV + image2.lens_FOV)
If ((Abs(image1.yaw - image2.yaw) + Abs(image1.pitch - image2.pitch)) < diff_threshold) Then assign the same stack ID to the two images.

Or should I manually declare both pics as a "stack" even though the overlap is only 50%?
I think manually specifying the stacks is the only approach given that your images aren't very well aligned.

Max

Gr8Scot
Posts: 5
Joined: Wed Apr 02, 2008 2:45 pm

Post by Gr8Scot » Wed Nov 04, 2009 1:05 pm

Ok, then that makes a lot of sense onhow my multi-focus panos are turning out.

So Max, on a risk that this isn't a question you've already been asked a lot, when you take a panorama with multiple focus', it sounds like you take multiple pics at different focus' of a scene before pointing the camera to the next part of the panorama. Is that how you do it?

I was taking a row of pics at one focus and then re-doing the row at a different focus. However, the images of the first run through the row did not line up all that well with the next run through the row with a different focus. The result was that Tufuse didn't seem to want to perform focus blending, probably because it didn't detect any stacks due to the por image registration from one pass through the row to the next.

Again, I hope this makes sense. I'm finding this a very difficult thought to put into words. But thanks for trying to understand what I'm asking and providing your thoughts on it.

By the way, every time I use PT assembler, I just keep thinking how cool it is. So much control and it works so well. I just love it!

-Scot

maxlyons
Posts: 3649
Joined: Fri Jun 20, 2003 8:55 pm
Location: USA
Contact:

Post by maxlyons » Wed Nov 04, 2009 2:09 pm

Gr8Scot wrote:when you take a panorama with multiple focus', it sounds like you take multiple pics at different focus' of a scene before pointing the camera to the next part of the panorama. Is that how you do it?
It depends on what equipment and camera I have. Yes, this is certainly one approach that I have used. I find it works best with my small digicam, where I can dial in specific focus settings using its on-screen menu. With an SLR, I find it tricky to alternate back and forth between multiple focus points precisely.

If I'm using a tripod head with click-stops (like the Nodal Ninja 5), then I shoot a row at one focus setting, adjust focus and then reshoot the row with the next focus setting. The click stops allow me to reposition the camera correctly when shooting the second row.
I was taking a row of pics at one focus and then re-doing the row at a different focus. However, the images of the first run through the row did not line up all that well with the next run through the row with a different focus. The result was that Tufuse didn't seem to want to perform focus blending, probably because it didn't detect any stacks due to the por image registration from one pass through the row to the next.
There are two separate issues here. The first ("images of the first run through the row did not line up all that well with the next run") is that due to the way in which you took the shots, the images aren't even close to being aligned. However, the second issue ("TuFuse didn't seem to want...") isn't related to the first.

TuFuse will attempt to perform focus blending on any images that you tell PTAssembler are part of the same stack. Even if the images only overlap by 10%, TuFuse will still try and focus blend that 10% overlap region. And, assuming you have optimized the project and nothing went wrong, then PTAssembler will be able to position the images before feeding them to TuFuse so that they are aligned. Regardless of how much (or little) they overlap, they should still be aligned correctly by the time PTAssembler hands them off to TuFuse.

Hope this helps.

Max

Gr8Scot
Posts: 5
Joined: Wed Apr 02, 2008 2:45 pm

Post by Gr8Scot » Wed Nov 04, 2009 8:58 pm

Yes, that helps quite a bit. I think I get it now. Thanks very much for helping on this. I can't wait until my next pano photo session!

-Scot

Castillonis
Posts: 14
Joined: Sun Jun 22, 2008 5:52 pm
Location: Portland, Oregon

Capture of Focus and Exposure bracketing for Panoramas

Post by Castillonis » Thu Nov 12, 2009 6:02 pm

If you want to automate the capture of focus bracketed or focus and exposure bracketed panoramas, there are some solutions that allow you to make this more precise and easier.

You can control your camera with Breeze Systems DSLR Remote Pro for Canon Cameras and NKRemote for Nikon cameras. You can automate a series of shots by controlling the Breeze Systems software with the freely available autohotkey software.

Canon
http://www.breezesys.com/DSLRRemotePro/features.htm
Nikon
http://www.breezesys.com/NKRemote/features.htm

So, If you wanted to mix focus and exposure bracketing you would exposure bracket at each focus for a particular position and then move to the next position in your panorama. Then you would combine the exposure for each focus for all of the positions. Then you would focus blend the results. Then you would stitch your panorama.

Now this is where you need to be able to make tradeoffs. If you were previously obsessed with making the largest panorama with a long lens, you need to be able to make tradeoffs for the particular scene to achieve the best image. So, start out with less ambitious objectives to learn and perfect this method. Look at the scene and think about which qualities would improve the overall image. Exposure? Focus? Time required to capture the shot? Is the light changing, or are things moving?

Some people are doing this with the Merlin / Orion telescope mount open source Papywizard software along with autohotkey. Though, because there are so many hurdles, only a few people are doing it.

You can also do 5 or 7 focus brackets with the Olympus 500 4/3" camera body. Unfortunately the newer cameras do not have this feature and the 4/3" sensor is slightly smaller than an APS-C size partial frame sensor.

The Canon point and shoots have minimal focus bracketing feature with 3 focus brackets. You will need to manually set the exposure brackets as the AEB only works in Av mode and not M mode.

maxlyons
Posts: 3649
Joined: Fri Jun 20, 2003 8:55 pm
Location: USA
Contact:

Re: Capture of Focus and Exposure bracketing for Panoramas

Post by maxlyons » Thu Nov 12, 2009 11:05 pm

Castillonis wrote:So, If you wanted to mix focus and exposure bracketing you would exposure bracket at each focus for a particular position and then move to the next position in your panorama. Then you would combine the exposure for each focus for all of the positions. Then you would focus blend the results. Then you would stitch your panorama.
Or you could just load all the images into PTAssembler and let it figure out which images should be exposure blended and which images should be focus blended and do the appropriate type of blending in one step.

By default, PTAssembler uses TuFuse for stacking, and one of TuFuse's advantages over other "fusion" programs it that it analyzes the input images and figures out which images ought to be exposure blended, which images ought to be focus blended, and does the appropriate type of blending.

Max

Castillonis
Posts: 14
Joined: Sun Jun 22, 2008 5:52 pm
Location: Portland, Oregon

Post by Castillonis » Fri Nov 13, 2009 5:44 am

Max, that is excellent that PTassembler figures which type of blending to apply. Though I think that this person needs to use a different method of capture to be successful. He needs to do the bracketing in exactly the same position and then overlap adjacent positions by aproximately 30% to yield good panoramas. This thread is in the Tfuse part of the forum, so maybe this is not a good place to discuss this.

dsjtecserv
Posts: 551
Joined: Sat May 05, 2007 3:25 pm
Location: Northern Virginia
Contact:

Post by dsjtecserv » Fri Nov 13, 2009 9:13 am

The original poster's question concerned the use of TuFuse within PTAssembler, to which Max has already responded. I find your suggestions interesting for non-PTA applications, and I think I might look into one of them, but it was actually your post that was the non-sequiter!

Dave

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest