FYI: Test Results Comparing Different Methods of Resizing Images

[Michael Rostkowski]

I posted this in alt.php, but as a reply, so now I'm posting it here for
people
who might have missed it but still could find it useful. Thanks to Andy
Hassall for
giving me good information.

I compared the smooth image resampling used in PHP to create good quality
thumbnails of large images, used commonly in image galleries. i compared
ImageMagick and PHP (with GD2) and compiled some statistics below.
I ran two test. Both used a set of 10 photos around 3MP in size. On the
first
test, I created two thumbnails for each photo, one 700 pixels in width and
another at 120 pixels in width. On the second test, I created just one
thumbnail, 120 pixels in width. Time was counted in "minutes.second s."
TEST#1 -- IMAGEMAGICK*
-----------------------------------------------------------
METHOD TIME IMAGE QUALITY
thumbnail 1.33 good
sample 0.24 bad
resize 1.33 good
size/geometry 0.42 good

TEST #1 -- GD2 (PHP)
-----------------------------------------------------------
METHOD TIME IMAGE QUALITY
Imagecopyresamp led 1.57 good
TEST#2 -- IMAGEMAGICK*
-----------------------------------------------------------
METHOD TIME IMAGE QUALITY
thumbnail 0.26 good
sample 0.19 bad
resize 1.13 good
size/geometry 0.06 good

TEST #2 -- GD2 (PHP)
-----------------------------------------------------------
METHOD TIME IMAGE QUALITY
Imagecopyresamp led 0.56 good
*ImageMagick Command Methods:
thumbnail : convert -thumbnail $destSize $sourceFile $targetFile
sample : convert -sample $destSize $sourceFile $targetFile
resize : convert -resize $destSize $sourceFile $targetFile
size/geometry : convert -size $destSize $sourceFile -geometry $destSize
$targetFile

CONCLUSIONS:
-----------------------------------------------------------
There were four different imagemagick commands I tried out in all. Overall,
the winner came out to be the imagemagick command. This
method beat out PHP by a good margin, creating thumbnails in about 1/3 the
time it took PHP with similar quality in the first test, and then taking the
cake
in the second test by creating thumbnails in about 1/10 the time as PHP.

Imagemagick seems to increase in performance compared to PHP when creating
a smaller thumbnail versus a relatively large thumbnail, as seen by test #2
versus test #1. This means that a site creating small thumbnails will see
even more of an advantage by using Imagemagick.

The fraction of the time it takes for imagemagick to create smaller
thumbnails makes it almost necessary to use for dynamically created
thumbnail galleries that need to trim resources.

MR