Re: [PanoTools-devel] PTmender

[Max L. <max...@ve...>]

>  Alan> an interesting pattern of failures. Binaries are from 
>  Alan> pano12-2.8.1-win32.zip on sourceforge. 

The PTMender code has changed a lot since this release.  I recommend compiling 
and testing the latest code from Sourceforge, or using a recent version that 
someone has compiled.  I suspect that a lot of the problems have already been 
fixed.  My own distribution of PTMender/Pano Tools (Windows only) is here:

http://www.tawbaware.com/maxlyons/pano12ml.htm

>  Alan> With exposure correction set to both colour and brightness all output 
>  Alan> formats work, although for my main project it produces some strange 
>  Alan> colouring in sky sections especially those with cloud.

This is the same as with PTStitcher.  The colour correction code was never 
really finished by Dersch (it was only released in "beta" form before he 
stopped working on Pano Tools), and does behave this way.  

>  Alan> Another possible problem is with handling large output images. 

The code to remap images to TIFF_m format works pretty reliably regardless of 
how much memory you have installed.  In order to create most other output 
formats, PTStitcher (and hence, PTMender) needs to load the entire output image 
into memory at least once.  So, if you are trying to create a 600 megapixel 
image, PTStitcher (and PTMender) will try and load the entire image into 
memory.  And, a 600 megapixel image requires 2.4 GB of free RAM (one pixel 
needs four bytes of memory), which is a problem on an operating system like 
Windows XP that limits the amount of memory any process can address to 2GB.  
PTStitcher (PTMender) asks the operating system for 2.4GB of RAM, and the 
operating system says..."no".  Then PTStitcher (PTMender) crashes because it 
can't get the memory it needs to load the image.

PTMender could be recoded.  All routines that require loading the entire image 
into memory would have to be reworked.  So far, nobody has done this.

On the other hand, PTMender can output arbitrarily large images in TIFF_m 
format, and Enblend can blend these into a final image, regardless of how much 
RAM you may or may not have.

> I am not very knowledgeable about the file limits of TIFFs and JPEGs
> but you might have hit one of those. 

A good point.  TIFF files can be no larger than 4GB in size (assuming a recent 
version of the tiff library was used for whatever software you are using) and 
JPEG has similar constraints as well.  So even if PTMender was recoded, these 
constraints would still exist.

>  Alan> One particular gripe in trying to test ptmender is documentation. The 
>  Alan> zip file contained little in the way of useful documentation. 

> I agree, but part of the reason is that it the same as PTstitcher :)

Agreed.  That was always a problem with PTStitcher...incomplete, out-of-date 
documentation.  

However, this is an excellent way for "non-coders" to contribute to this 
project...writing a user-friendly guide to PTStitcher/PTMender.  In fact, "non-
coders" would be uniquely positioned to write such a document, given that they 
are more likely to understand the difficulties encountered by "non-coder" users 
who can't/won't look at the source code.  

Armed with the existing documentation (sources below), some time to do some 
trial runs, and some patience, I've no doubt that someone with Alan's writing 
skills could contribute some excellent documentation.  This would be an 
extremely welcome contribution to the project.

The most updated script reference that I've seen for PTStitcher is here:

http://panotools.cvs.sourceforge.net/panotools/libpano/doc/stitch.txt?view=markup

Ben Kreunen has some good documentation for the stitcher script syntax here.  
It is a little old, but still very relevant:

http://www.path.unimelb.edu.au/~bernardk/tutorials/360/readme/index.html

And, I've described the new "PTMender specific" script parameters here:

http://www.tawbaware.com/maxlyons/pano12ml.htm

Max