>>So, we have 500,000 cards (250 card - two sides).
>>
>>I will have to name the card by id+a and id+b (100001a, 100001b, 100002a...) for the both card sides.
I'd rather go with a meaningless filename, maybe prefixed with the user ID, something like 100001_0001.jpg etc, because it starts with two pages per person, but once this is up and running, there may be others.
>- archive the original scan bitmaps at full resolution
>- process them into .JPGs or .PNGs for production use
>
>Storage estimates:
>24-bit colour scans @ 300dpi: ( 5.6 * 300 )* ( 4.3 * 300 ) * 500,000 * ( 3 bytes/pixel ) = 3.25 * 10^12 bytes, roughly 3TB.
>Same parameters but 1-bit B&W would be 1/24 the size, or about 135GB.
For handwriting and documents in general, 150-180 dpi should be enough; add to that the low number of colors for the signature - b/w would suffice - and the compression, we may reduce this to maybe a factor of about ten. Nevertheless, I'd also recommend ample storage, because the size of the signature scan can depend on many parameters of the scan - contrast, black-gray-white threshold settings etc, so maybe we can assume compression no better than 50%.
> Try not to store too many images in any single folder (I seem to recall hearing that Windows doesn't like more than about 1,000 files in a folder).
That was in the olden days when DOS was the underlying OS and FAT was the filesystem. Nowadays, NTFS is supposed to handle this much better - and I've had a directory with 15000 images without much of a problem. However, speed may be an issue; directories being indexed in some System Volume Info (aka volume.sys or system.dir on VAX 22 years ago, IIRC ;)
>You may want to post-process these bitmaps into smaller .JPGs or .PNGs that are optimized for a particular use. For example, you may want to display thumbnails on-screen @ 72dpi, or display/print just a signature block.
Quite simple to do on-demand with GDI+ (the one by Cesar Chalom or the other by Alexander Golovlev), i.e. if the small image isn't there, create it when needed but not in advance (too much processing, and for what - who will ever look at all 250000 thumbnails?).
>However, with the large number of scans to do, you're looking at high-volume scanning. You will probably want to investigate the image-processing capabilities of whatever high-volume solution you decide to use.
We can expect increased employment in his neck of the woods.