Batch-combining large images

Question

I have a bunch of equal-height images in PNG format that I want to arrange horizontally into a single image, without any overlapping. I need a software that can do this either automatically (e.g. with a batch file) or with only a few mouse clicks. Neither XnView nor IrfanView seem to have this option in their batch processing repertoire. ImageMagick's montage/convert tool can do this, but it's extremely memory-inefficient and uses over 10 GB RAM for combining images with only 0.5 GP in total (at 24bpp) which is unacceptable. DipStych has a similar problem. And from what I've seen, panorama stitching software such as ICE generally can't combine images without overlapping them.

In theory this should be a trivial task: Load all images into RAM, concatenate the arrays, save to disk, done. Which software can actually do this?

Answer 1

You already mentioned imagemagick (IM). It should just work with bigger files, too. Just try it. This software by default allocates a seemingly sensible amount of memory depending on available hardware resources to facilitate fast processing of operations in main memory and not being dependent on disk access times.

First thing for you would be to use 8bit instead of 16bit colordepth version of IM (or alternatively graphicsmagick, which often works even quicker).

If you definitely want to put IM (or gm) in chains and wait longer (orders of magnitude) for completion of the task, you might want to try their "-limit" option (https://www.imagemagick.org/script/command-line-options.php#limit): as soon as relevant proposed limits are exceeded, the image pixels get allocated to disk instead of main memory.

Answer 2

You can use the Hugin tool chain as if you are stitching scans together as described at http://hugin.sourceforge.net/tutorials/scans/en.shtml where the Advanced Techniques section talks about scripted usage. You will have to do some digging into the documentation but you can force the control points, (used for aligning images), to be on the common edges of the adjacent images.

Hugin is a free, (gratis & open source), cross platform, tool chain for doing panorama stitching, focus & exposure stacking, correction & more.

Note on memory usage:

JPEG images are a compressed format so when considering memory usage any operation, (editing, converting, stitching, etc.), will use at least width pixels x height pixels x bytes_per_color x number of color channels bytes of RAM. Thus an example file I have picked at random is a jpeg with:

• Size: 6,334,976 bytes on disk (call it 6 MB)
• Image Size: 4896 w x 3672 h = 17 978 112 pixels
• Mode RGB & Bits/Channel 8 = 3 bytes per pixel
• This gives a minimum RAM requirement for the image data alone of 53,934,336 - (near enough 54 MB) for the single input image

If you are generating a new image you will need, at minimum, enough space for the input image(s) plus that for the final output image so merging 2 photos like that above will need at least 4*54 MB = 216 MB to merge just two "6 MB" Image files.

Answer 3

long late reply, many games take a bunch of small little pictures. and toss them all into a program. that then spits out a single file. with x,y (top/left corner) and x,y (bottom,right corner) within the single file of all the pictures within the single file. and then they just reference the array of which picture to display. most likely used for animation simple animation purposes. of say gun fire (little explosion) but without the gun. gun itself, unit standing in each direction, etc... each little image generally has a transparent background, with just the needed pixels to be displayed. (diff) in essence. so overall single file is a bit more smaller. and the game programmers able to load single file into memory taking up less space. the individual little pictures can by any size / shape for most part. it is the array list it puts out with the single large file. that is important.

working with larger images on other hand. you might look at "tif" file/picture format, or look at pure file reading/writing/appending to file. without loading them all to memory. read file. write to file, read file, append to file, read file, append to file. so mim amount of memory is used.

html5 -> canvas and some javascript might work. run a multi file upload button via html5 file upload with multi option. and a javascript that keeps expanding left or right adding images next to each other onto the canvas element, and then save the canvas via download button. or auto prompt with a "download" window in selecting were to save.