Batch detecting quality in large number of images

[giangstar]

I have a large batch of images (in the tens of thousands). They are of varying quality. Some are 300 dpi, 4000 pixels by 6000 pixels. Others are 150 dpi, 4000 pixels by 6000 pixels. The sources of these images vary as well. File sizes range from 5MB jpeg files to 100 MB jpeg files. I can tell that even with some of the large jpeg files, the images were enlarged so that it seems like it would be good to print,. But, in fact, when you go into the image in Photoshop, the resolution just isn't there. How do I detect this for a large number of images?
Thanks.

[McQ]

First, there's one thing we need to make clear: a file's image quality is entirely independent of it's DPI setting. Unless you have done substantial resizing, or if the photos were shot with many different cameras, you should be looking solely at the resolution in pixels to assess quality. DPI only comes into play when it is used to determine the size of the print.

Just thinking out loud here...I cannot think of an easy solution:

Since many of your files have been resized and are from different cameras, I would first try to group them based on EXIF information. You could place all of the edited versions in a different folder or give the filen ames a suffix. You could also group these by camera model. There's a number of software programs which will sort photos based on EXIF, which will help make this a little more automated. However, it will still require a lot of intervention...something I think you'll want to avoid with tens of thousands of images.

Another option, which is very crude and brute force, would be to sort the files based on the ratio of their file size to their megapixel count. A higher ratio would mean that in general, the image contains more detail. However, I would guess that some of these files might use different levels of JPEG compression, and some photos might contain smooth skies and little texture or noise (creating a smaller JPEG file at the same resolution) -- hence the preface of this technique being crude and brute force. This would also only work when comparing files with similar megapixel counts, so would likely also not help with what you're trying to do.

Overall, it really depends on what you're trying to do. Would you really ditch a photo because it does not have as much resolution? Are you trying to assess which of many similar photos will make the largest and most detailed print? Are you just trying to figure out which ones to make into large prints and which to save for digital only viewing? All things to consider when sorting the photos...