TuFuse - anyone tried it?

Has anyone here ventured into using either TuFuse (which also comes packaged with PTAssistant) or enfuse? They are both freeware, command-line tools which do automated exposure and focus blending.

Of the two, TuFuse appears to be the more advanced in development; it features a two-stage process with auto-detection, separating images of like exposure into groups for focus blending, then combining the resulting images using exposure blending.

The program is much faster than HDR and doesn't require nearly as much tweaking. It also seems to beat out manual exposure masking in most cases.

The image above was processed using HDR; it took me a significant amount of time to tweak to get the sun and shadow to my liking.

And this is an image produced with TuFuse. I converted the RAW images into TIFFs as-is using ACR, then ran TuFuse with its default parameters. I brought the image back into PhotoShop, re-added the AdobeRGB color profile, adjusted the color balance slightly toward yellow to come "close" to the reference HDR, and adjusted master saturation my "standard" amount. Total processing time, maybe 15 minutes.

Notice the better sky coloration and more defined sun rays. The only thing I don't like about this version as much as the HDR version is the blurred ripple - TuFuse decided that the more blurry ripple was better exposed.

Looks cool, how do you use a command line tool?

For Windows users, you usually go to the Start menu and select 'Run...' then type in 'CMD' and press Enter. That brings up a Windows (DOS-like) command window. From there, at least for this, you change directory into the directory containing the TIFF or JPEG files you wish to blend, and enter the command... E.g. I entered the following:

cd C:\Photos\TestFolder
tufuse -o output.tif file-2stops.tif file-1stop.tif file+0stops.tif file+1stop.tif file+2stops.tif

I had previously copied the TUFUSE.EXE file into that directory (I just got it and wanted to test. You can put the executable file into a permanent directory and update your PATH environment variable, or copy it into, say, C:\Windows\system32)

If you don't like command-line, there is a Droplet (a small drag-n-drop Windows application) that comes with enblend, the parent project of enfuse, and EnfuseGUI which allows you to use enfuse without also using enblend (enblend is a panorama stitching blending tool.)

Enfuse has the advantage of having working Linux and Mac OS X ports; TuFuse has the advantage of automated determination of whether you're trying to do focus blending or exposure blending. Both work the same way - they perform weighted comparisons on a pixel-by-pixel basis, comparing local contrast, saturation, and exposure. By changing the weighting, you change the behavior of the program. Weighting for contrast gives you focus blending; weighting for exposure gives you exposure blending.

BTW, both of these are relatively "raw" right now; I expect in the near future to see a similar tool made available as a PhotoShop plug-in. (In fact, Helicon Focus probably uses a method very similar to these tools to do its focus blending...)

Holy cow Les, thanks for this! I'm a long time PTAssembler user, but I've been hanging out a lot less on that forum since joining here. I have the greatest respect for Max Lyons and his software, so I'm off to check it out.

Thanks for sharing this, Les. I'm a long time PTAssembler user too, but I haven't been over to Max's site in a while.

Too funny Les, I have been testing both of these in the past few days. Tufuse and Enfuse are essentially the same, but Enfuse adds a very critical option; it allows an alpha mask to be added to the images.

This may seem minor, but it's critical to avoid motion induced blurs in multi-expose image sets. This also allows you to fine tune your blends if you want some object to definitely come from one and only one image.

The next version of Hugin will incorporate the Enblend/Enfuse/HDR capabilities, so pano makers will have a fully integrated tool set once that happens. If you don't mind beta versions, you can already test this out.

Craig

I didn't even know about these until I saw an article link on Slashdot's sidebar on exposure blending in Linux. I figured it was an article on HDR in Linux and was pleasantly surprised that it was much, much more.

Thanks for the additional info, Craig.

...You can put the executable file into a permanent directory and update your PATH environment variable...

Holy bat shiat, is this English? Being a non-genius at computers, the room started spinning as I read this whole description. Gotta be a techie to understand it.

Hence the GUI interfaces. Command-line for photo processing is, um, not ideal.

Between Hugin, the Enblend Droplet, EnfuseGUI, and PTAssembler, there's a decent selection of GUI tools to help you in your quest for better blending.

When I come up for air in the next few days, I'm going to try this on a couple of shots that have completely failed to work via HDR or manual blending and see just how much of a miracle-worker this thing really is...

Wow Les this is amazing, thanks a lot for sharing,

Bill

One question, how do you align your images? Or more correctly make them the same magnification when they are overlaid on each other. I find when doing focus blends that the object size (zoom) (eg a flower) changes with focus point (even when tripod mounted etc etc), its just the way that the lens works. So how do you adjust the object "sizes" so that they will blend together correctly and give a sharp focus blend?

Thoughts?

Bill

Hello...I'm the author of TuFuse.

Tufuse and Enfuse are essentially the same, but Enfuse adds a very critical option; it allows an alpha mask to be added to the images.

TuFuse does this as well. Pixels that are masked (i.e. the alpha channel is zero) are not used when performing image fusion. Pixels that are not masked (i.e. alpha channel not equal to zero) are used during fusion. You can edit your images in your favorite image editor to adjust the mask to inclue/exclude regions as desired before invoking TuFuse.

The next version of Hugin will incorporate the Enblend/Enfuse/HDR capabilities

The current version of PTAssembler already has this. In addition to supporting Enblend and Enfuse, It also supports TuFuse and CombineZM (focus blending) for stacking, and a variety of other programs:

http://www.tawbaware.com/ptasmblr.htm
http://www.tawbaware.com/ptasmblr_help_stack_example.htm

One question, how do you align your images? Or more correctly make them the same magnification when they are overlaid on each other. I find when doing focus blends that the object size (zoom) (eg a flower) changes with focus point

PTAssembler can align images and act as a sophisticated "GUI" for TuFuse (and other programs). In fact, PTAssembler's new "auto-align" feature (consult the PTAssembler documentation for details) is specifically designed to align a stack of images where the focus has been adjusted between frames, which causes a change in object size (or image field-of-view [FOV]).

Hope this helps,

Max

TuFuse does this as well. Pixels that are masked (i.e. the alpha channel is zero) are not used when performing image fusion. Pixels that are not masked (i.e. alpha channel not equal to zero) are used during fusion. You can edit your images in your favorite image editor to adjust the mask to inclue/exclude regions as desired before invoking TuFuse.

Max, glad you dropped by. I tried the same images in Enfuse and Tufuse, all of which had alpha channels in them. The Tufuse and Enfuse output images were vastly different with Enfuse being what I wanted and expected based on the alpha masks.

Do I have to do something special to get the masks used correctly in Tufuse? Also, I assume that a mask with grey levels (ie. between white and black) would cause Enfuse/Tufuse to scale back that corresponding image pixel by the amount of the grey level.

So for example, if the mask is at 45% then that image pixel's weight would be scaled back by 45%. Is this true?

Thanks,
Craig

if the mask is at 45% then that image pixel's weight would be scaled back by 45%. Is this true?

No...as far as TuFuse is concerned a pixel is either unmasked or unmasked. If the alpha channel is non-zero, then the pixel is considered to be "unmasked". If the alpha channel is zero, then the pixel is considered to be "masked" and is not used.

Max

No...as far as TuFuse is concerned a pixel is either unmasked or unmasked. If the alpha channel is non-zero, then the pixel is considered to be "unmasked". If the alpha channel is zero, then the pixel is considered to be "masked" and is not used.

This must be the cause of the differences I see between Enfuse and Tufuse then. The masks definitely have everything from white to black, so Enfuse must be taking that into account while Tufuse is not. Any chance you will be updating Tufuse to use a mask as I described?

Thanks for the information,
Craig

Wow, this is a focus blend I made after listening to the advice above, amazing, the stamen is 4 inches long, this image might not be possible without this cool software!

thanks,

Bill

Any chance you will be updating Tufuse to use a mask as I described?

I hadn't planned on it. I'm not sure how generally useful this would be, and I have a list of other features I want to implement first.

Max

I've updated TuFuse to version 1.30.

This version adds several new features including support for JPEG and RAW files, more configurable output choices (bit depth/compression) and more flexible "auto-bracketing" choices. Here is a complete list of changes.

Max