reasonable PHP memory_limit for image resize

Question

I want to enable users on my site to upload images to their accounts. The images get resized into 4 different sizes required across the site.

I have been using Pear Image_Transform but I kept getting "bytes exhausted" fatal errors on certain types of jpgs (all files tried under 2mb). So I moved to a dedicated server with Pentium Dual-Core E5200 @ 2.50GHz and 2GB ram. Uploaded the same image resize code - same error. I upped the RAM in php.ini to 64M but site get the same problem on certain types of jpg. Also tried wideimage class - same error (error is always with imagecreatefromjpeg()). (Using GD2). All works fine locally on my mac.

Is this really a memory issue, what's a reasonable memory_limit for my set up + image resizing?

Answer 1

It seems that ini_set("memory_limit","256M"); has covered it. Also, using memory_get_peak_usage() I discovered that my image resize (with problem jpgs) were using in the region of 90mb for a 1.8mb image.

Also of interest is http://uk.php.net/manual/en/function.imagecreatefromjpeg.php#64155 for a way to dynamically assign a memory limit (which I have yet to try).

Answer 2

90Mb of memory for a 1-2 Mb jpeg seems odd. I haven't seen your code, but perhaps you're opening the file multiple times for each size instance? Try opening the file once, (resize it, save it) X 4, then close it.

I currently have a site where users upload images, and not running into an error, my settings are 16Mb for PHP scripts and 2Mb file upload limit and I use this library instead of pear

code example

$image
->open('uploaded.jpg')
->resize(500, 500)
->save('large.jpg')
->resize(100, 100)
->save('medium.jpg')
->resize(25, 25)
->save('small.jpg')
->close();

Answer 3

A rough guide to how much memory you're going to need can be calculated like this

$imageInfo = getimagesize( $sourceImagePath );

// a check to make sure we have enough memory to hold this image
$requiredMemoryMB = ( $imageInfo[0] * $imageInfo[1] * ($imageInfo['bits'] / 8) * $imageInfo['channels'] * 2.5 ) / 1024;

This is quite rough and includes a fudge factor, 2.5, which you may want to experiment with.