large number of tiff images and files

[mmuller]

#!/bin/bash
#There are approx 10 tiff files with each image/scene/frame of 190x190 16bit-pixels. There are approx 30000 images per file.
#Plan: get center slice from each of the 30K images and concatenate all these slices into 1 image file. So will get 1 new image file per multi-image.tif file.
#pseudocode
#loop
# get multiimage.tif filename
# get number of images in file (nImages=identify...)
# loop 0,nImages
# make a filename with # to use for crop output
# get slice out of middle of current image and put in numbered file
# endloop - get next image in multiimage.tif
# Now have $nImages slice-files in tempDir and want to combine them all
# into 1 image file. There will be 1 new file per muti-image.tif file.
# Use convert to combine all ~30K slice-files into 1 new image file.
# Remove tempdir/* (all slice-files).
#endloop - get next multiimage.tif

mkdir "tempDir"

for file in imagedir/*.tif; do

nImages=`identify -ping -format %n $file`
fileonly=$(basename $file)

for (( i = 0 ; i < $nImages; i=i+1 )); do

orderFile=tempDir/$fileonly`printf "%06d" $i`.tif
convert -quiet -crop "1"x+"95" $file[$i] $orderFile

done

convert tempDir/* +append ${fileonly}_newfile_bydir.tif
rm -f tempDir/*

done

There are basically 3 commands. Can any of them be sped up?
1 is posted as bug report, 2 gets run 30K times to produce 30K files, 3 appends 30K imagefiles.
1) identify
2) convert -crop
3) convert *.tif +append

Here is an example tiff file with 29516 images in it (0-29515). It is 2.1GB ungzipped.
ftp://ftp.swri.org/pub/incoming/multi29 ... bit.tif.gz

I have 2 alternatives to 1) identify. tiffinfo requires installing another package. They are:
alt1)
#!/bin/bash
convert -quiet multi29515images16bit.tif[0] delme.tif
onesiz=`ls -ln delme.tif | awk '{print $5}'`
allsiz=`ls -lnH $file | awk '{print $5}'`
rm delme.tif
nguess=`expr $allsiz / $onesiz`
nlow=`expr $nguess / 1000 \* 1000`
nhigh=`expr $nlow + 1000`
ncount=`identify -format %n -quiet multi29515images16bit.tif[$nlow-$nhigh]`
nImages=`expr $nlow + $ncount`
nImages=`expr $nImages - 1`
alt2)
tiffinfo multi29515images16bit.tif 2>errs | grep TIFF | wc -l

convert -version
Version: ImageMagick 6.5.4-8 2009-10-24 Q16 OpenMP http://www.imagemagick.org
Copyright: Copyright (C) 1999-2009 ImageMagick Studio LLC

I have a speedup for 2) the convert section. It only works for i=i+1, that is, when doing ALL the images. If skipping images, not shown but would be something like for(;;i=i+$skip), then the code below would not be appropriate. Here is replacement which does 1000 images at a time.
On my computer, it takes 6.7 secs to do the 1st 1000 images and about 4minutes for all 29K instead of the 9+ hours of the original code.

for (( i = 0 ; i < $nImages; i=i+1000 )); do
j=`expr $i + 999`
convert -quiet -crop "1"x+"95" $file[$i-$j] tempDir/${fileonly}%06d.tif
done

[fmw42]

Only one minor speed up that I can think of. Add -ping to your identify (depending upon what you are trying to get), it will help if it is not image statistics, but filename or width or height which are in the header. Thus the whole image does not need to be read. See -ping at http://www.imagemagick.org/Usage/basics/#controls