>>Think you can identity the image you receive with CRC ( I'd rather use MD5 or other method)
>
>Yes, the original image could be used to detect if it is the same or not. It is just that the CRC32 stored in the database is after the resize.
>
>What about MD5?

It's longer - 16 byte vs 4


>
>>And on an afterthought - do you need to store the resized image ? Can you resize it just for display purposes when you need it in a certain height and width ?
>
>Yes, this is like that for various reason. 50 million files will take a much bigger disk space if not resized. Also, this is for performance access. If the resize is done in real time, it will not be a fast as otherwise. Also, this is batch display. So, when displaying a lot of images, if they are resized, this will be considerable.
>
>>Don't know which business it is - but could you throw away the previous one if they send you another ?
>
>But, for that, I'd need to know if this is a duplicate or not.


Why not do the checksum on the one you receive ? If you get another one and the checksums are the same, then assume they are identical
( or just overwriting it is another possibility )

After calculating the checksum, resize it and store it ( but do not calculate any checksum on a resized one)