Trying to understand losses in jpg compression

Question

I have a bunch of jpegs taken with a Nikon D3200 at 6016 x 4000. They weigh between 9M and 12M each.

To try and reduce space a bit, I compressed copies with the default Windows utility, and the new images are 3411 x 2268 and weigh between 0.4M and 0.5M. That is, there's been a 20:1 reduction in file size, while pixelwise the reduction is 3:1.

I look at the images in my screen, even zooming several times, and I struggle to see a difference in quality; I'm aware than I'm no expert, though.

So my questions are, what is a more normal rate for jpeg compression: 6016 x 4000 at 10M, or 3411x 2268 at 0.5M? Is there anything I should be worried that I'm losing from the original files?

Answer 1

The fact that should be realized is that your 6016 x 4000 pixel images are 72,192,000 bytes size before compression. That is roughly 69 MB, and is 3 bytes of R,G,B image data per pixel. That is simply the size of your RGB image data, 69 MB (uncompressed). That is 6016 x 4000 x 3 bytes per = 72,192,000 bytes of image data in every image. That is simply the size of your image data.

Then JPG is drastic compression, not lossless, meaning, the images do not decompress to be the same quality as before compression. The more compression, the worse they get. Here is an idea at my own site about that problem: https://www.scantips.com/basics09b.html

Your 9 to 12 MB compression is the "normal" reasonable size you ask about. That is already perhaps around 1/8 of the actual data size, which is already a tremendous reduction. That would be the cameras highest quality choice called Fine (Fine JPG Quality, as opposed to less quality). Your camera manual says 11.9 MB is the D3200 average typical file size. (different scenes can compress a little different, but in this ballpark.)

It would not seem reasonable to Not want high quality. Just use your cell phone if you don't want the best. :)

You can make image files smaller by resampling to fewer pixels, but still Fine JPG quality. That is camera Large, Medium, Small, which is just resampling. Resampling to 1/3 size in pixel dimensions is 1/9 size in area, and in bytes. That could be about right for your 1920x1080 video monitor, unless you needed to crop it substantially. Nothing really wrong with that except for printing prints larger than 6x4. But when you end up with the photo of your lifetime, you might wish it larger.

Or you can use extreme JPG compression, and pretty much ruin your images. Both resampling and JPG compresion are irrecoverable, other than going back to retake the picture.

Your 0.5 MB images should show their own faults. I hope you just made copies, and kept the original files too.

Answer 2

Jpeg is a compression-based image algorithm. Very roughly, instead of storing full color information for each pixel, it saves space by saying "hey, the next six pixels are all the same color". Bumping up the compression level makes this happen more often, despite the pixel colors not quite matching.

How much is too much? That depends on your needs (are you putting it on a website?), but mostly, when it no longer looks good - you've gone too far.

---

Separately, though, jpegs are lossy (when you make modifications, the image quality inherently degrades), so if you are concerned about keeping image quality, you should store the raw photo and use a raw image editor, like Lightroom.