JPG vs PNG.

[fotoobscura]

Afternoon.

I archive all my scanned images to TIFF but their file sizes are unwieldy for cloud storage and for casual viewing.

Over the last few years I have been trying to figure out what image size/resolution I should convert TIFFs to for casual (but functional) viewing. By that I mean a size that is still printable at a fair size (say, 16x20in) but not a huge file. PNG's are (typically) for me 90% the size of an LZW compressed TIFF, not much of a savings. A JPEG at 100% quality is roughly one third the size of a PNG. Now that's quite a savings.

The question is- Am I losing quite a lot of detail/color/resolution comparing a 100% quality JPEG and a compressed PNG? If so, is there a method to determine how much that is? Or is this hypothetical? If I'm losing say 10% "quality" on a JPG as opposed to a PNG, I'm willing to take that loss if the file size is three times smaller. Personally I cannot visually perceive any *quality* difference between a 100% JPG and an uncompressed PNG.

In my ultra unscientific study I have attached two images magnified roughly 700%. The first is a jpeg, the second is a png. These come from 2000dpi 35mm scans off of a Nikon LS8000 scanner @ 2000 DPI, compressed TIFF. The latter image (the PNG) has more color noise than the JPEG. I actually prefer the JPEG in this example!

Anyone out there with expertise to field this question?

Thanks in advance!

(all trolling >> dev/null)

 

[gmikol]

For viewing and printing, I don't see any problem with using a maximum quality JPG. You could even try inching the quality lower to see where there is a good size/quality trade off *for viewing and printing*.

For pixel-level editing, I would recommend sticking with TIFF. Repeated edit/save cycles could start to compromise the quality of a JPG. A workable compromise might be JPG --> TIFF working intermediate --> JPG Final

I think the JPG and the PNG in your example above look identical. Also, keep in mind that PNG and TIFF compression is lossless, whereas JPG compression does actually reduce information (though at maximum quality, that loss is quite small). Also, 16-bit-per-channel TIFFs don't compress very nicely, but 8-bit TIFFs do.

Using Photoshop or GIMP, there's a way you can evaluate the differences for yourself. Take the original TIFF or PNG, and save a copy as a JPG (at whatever quality you desire). Place the JPG copy as a layer above the original and set the layer mode to subtract. The image should turn black. For 2 identical images, it should be perfectly black. But if you use the equalize or curves/levels to stretch things out, you should start to see tiny differences. Once you know where to look, you can set the layer mode back to normal and toggle between the 2 images to see if there is actually a visible difference.

I hope that description makes sense.

Having said all that, I'll also add that storage is so cheap (3TB for US$130), why not just keep the original TIFFs and not worry about it?

--Greg

 

[fotoobscura]

Thanks for the reply.

Yes I'm aware of JPEG not being lossless but at 100% I think the differences *may* be visibly trivial.

I wasn't clear about the reasons for trying to determine this. My process is the following:

develop > scan to tiff > copy to two hard drives > convert tiff to 100% jpeg, upload to cloud storage > work mostly off of jpegs unless editing/digital neg/large format printing is needed.

The image comparing is great advice, I'll try it out.

Thanks!

For viewing and printing, I don't see any problem with using a maximum quality JPG. You could even try inching the quality lower to see where there is a good size/quality trade off *for viewing and printing*.

For pixel-level editing, I would recommend sticking with TIFF. Repeated edit/save cycles could start to compromise the quality of a JPG. A workable compromise might be JPG --> TIFF working intermediate --> JPG Final

I think the JPG and the PNG in your example above look identical. Also, keep in mind that PNG and TIFF compression is lossless, whereas JPG compression does actually reduce information (though at maximum quality, that loss is quite small). Also, 16-bit-per-channel TIFFs don't compress very nicely, but 8-bit TIFFs do.

Using Photoshop or GIMP, there's a way you can evaluate the differences for yourself. Take the original TIFF or PNG, and save a copy as a JPG (at whatever quality you desire). Place the JPG copy as a layer above the original and set the layer mode to subtract. The image should turn black. For 2 identical images, it should be perfectly black. But if you use the equalize or curves/levels to stretch things out, you should start to see tiny differences. Once you know where to look, you can set the layer mode back to normal and toggle between the 2 images to see if there is actually a visible difference.

I hope that description makes sense.

Having said all that, I'll also add that storage is so cheap (3TB for US$130), why not just keep the original TIFFs and not worry about it?

--Greg

 

[OzJohn]

Max or even slightly lesser quality JPEGS (say 10 in PS) do it for me most of the time. I detest TIFFS with a passion because of the file size and would rather use PSD if I don't want to lose anything. The key with JPEGS is to only save them once - work in some other format then convert when you are done. Do you actually print much at 16x20 or larger? Saving everything at that size seems a bit overkill to me if storage is an issue unless you have a particular reason.

gmikol: Great suggestion about the overlaid files - just shows that no matter how long you've used PS there are techniques out there that you would never have thought of until someone else spells it out. OzJohn

 

[fotoobscura]

TIFFs are necessary for high resolution scans (I regularly scan negatives at 4000dpi). LZW compression isn't the best but it works fairly well (at least in contrast to uncompressed TIFFs!). I seldom print at 16x20 or larger but I always want the option to do so with any available image.

It's not so much about storage is it's about speed. Browsing 50+mbyte TIFFs on even a fast computer can be torture, even with thumbnail caching. A JPEG will load up (to view) in a fraction of the time it takes to display a TIFF.

Best,

Max or even slightly lesser quality JPEGS (say 10 in PS) do it for me most of the time. I detest TIFFS with a passion because of the file size and would rather use PSD if I don't want to lose anything. The key with JPEGS is to only save them once - work in some other format then convert when you are done. Do you actually print much at 16x20 or larger? Saving everything at that size seems a bit overkill to me if storage is an issue unless you have a particular reason.

gmikol: Great suggestion about the overlaid files - just shows that no matter how long you've used PS there are techniques out there that you would never have thought of until someone else spells it out. OzJohn

 

[chuck94022]

My two cents. For the difference between jpeg and png, I would go jpeg. But as others alluded to, jpeg should be considered a final form, not an archival form for future re-editing. In other words, think of the jpeg as a print, and a tiff as a negative. You can re-edit the tiff, but you will introduce image degradation if you start re-editing jpeg. If the only thing you save is the jpeg, and toss the tiff, you have effectively destroyed your electronic negative.

I don't see a lot of reason to save to .png versus .jpeg. Yes, jpeg is lossy, but it is lossy like CD music is lossy. (And prints are lossy too - they don't represent everything that was on the negative, ever.)

Also, save the original film, it is your ultimate backup, and beats any electronic/cloud/etc. storage. I can't tell you how shocked I was the day Apple announced the end of life of iDisk, and gave everyone just a few months to evacuate. Luckily I didn't have much there, but geez. The cloud, so far, isn't an archival medium.

After the apocalypse, we film shooters will be the only ones with images, after all... ;-)

 

[fotoobscura]

What does JPG Quality Losses Mean?