You can transfer EXIF information, including the rotation info, with either exiftool or ImageMagick into a thumbnail, to let other software know how to display it correctly.
If you want to rotate a jpeg yourself (which is easily done with "magic in.jpg rotate 90 out.jpg" and, of course, other software) but lossless, there's https://jpegclub.org/jpegtran/ which accomplishes it.
When generating rotated thumbnails (without caring for "lossless") the best strategy in my experience is to first rotate the large original image and then generate the thumbnail afterwards, because image resizing will most often hide any rotation artifacts anyways.
The rexif crate supports editing, so you can apply rotation when resizing, and then remove the rotation tag from the EXIF data. Keeping EXIF isn't necessary for small thumbnails, but could be desirable for larger versions of the image.
I'm sure there's very good reasoning but I've never been able to convince myself of it - why does EXIF transform metadata exist? If you're taking a selfie, why doesn't the camera itself perform a simple rotation or mirror of the actual image data rather than using EXIF?
One reason I can think of: when you rotate an image in the windows photo viewer (and probably lots of other apps), it stores that in the EXIF metadata. So the rotation is lossless, by not having to re-compress the JPEG.
I could also imagine that the earliest digital cameras wouldn't have had the processing power, or RAM to store the entire image in memory after reading it from the sensor, in order to do such a rotation. Hence EXIF rotation as a cheap alternative.
You can rotate (in 90 degree increments) and mirror jpgs losslessly. I assume it doesn't require much ram or compute since you're just reordering the blocks.
Having read that article, I'd say this is not really an honest takeaway. I mean, I guess it is lossless, it just really flies in the face of what I'd consider reasonable baseline expectations.
They offer three "solutions" to the issue of partial blocks:
- cut them off and stop worrying about it
- show whatever the original JPEG encoder put there and stop worrying about it
- replace whatever the original JPEG encoder put there and stop worrying about it
I cannot imagine anyone honestly considering these and then shipping them. It's fine if the user manually consents, but automated, this is pretty objectionable. Extending the image in arbitrary ways or cutting off of it is really not what I'd expect from a codec to do when I ask it to perform a lossless transformation, particularly specific ones, like flip and rotate, which I'd expect to be outright periodic. And periodicity is violated immediately with the dimensions changing.
I think it does this after the photo has already been stored as JPEG. I think users would expect the image be persisted the moment you capture it, so any transformations after that are better done in metadata. When the image has been saved as JPEG the original image data is discarded so you can't transform without deteriorating the image.
A quick search suggests to me that it's only a lossless process if the image dimensions are a clean integer multiple of 8 or 16 (as the blocks can be 8x8, 8x16, 16x8, or 16x16), otherwise the edges must be reencoded. Never written a JPEG codec though, so happy to be proven wrong.
I too have always found this strange but I have never found a solid reason. Having an explicit declaration of intent is not a terrible feature in a data type so this is not exactly a bad choice. Also this means that a camera roll from a single sensor can be stored as an array in C since every image has the exact same size. It also does make life easier for the camera developers (but harder for photo viewer developers). If anyone knows the history here I would actually really be interested.
If you want to rotate a jpeg yourself (which is easily done with "magic in.jpg rotate 90 out.jpg" and, of course, other software) but lossless, there's https://jpegclub.org/jpegtran/ which accomplishes it.
When generating rotated thumbnails (without caring for "lossless") the best strategy in my experience is to first rotate the large original image and then generate the thumbnail afterwards, because image resizing will most often hide any rotation artifacts anyways.
I could also imagine that the earliest digital cameras wouldn't have had the processing power, or RAM to store the entire image in memory after reading it from the sensor, in order to do such a rotation. Hence EXIF rotation as a cheap alternative.
https://www.betterjpeg.com/lossless-rotation.htm
They offer three "solutions" to the issue of partial blocks:
- cut them off and stop worrying about it
- show whatever the original JPEG encoder put there and stop worrying about it
- replace whatever the original JPEG encoder put there and stop worrying about it
I cannot imagine anyone honestly considering these and then shipping them. It's fine if the user manually consents, but automated, this is pretty objectionable. Extending the image in arbitrary ways or cutting off of it is really not what I'd expect from a codec to do when I ask it to perform a lossless transformation, particularly specific ones, like flip and rotate, which I'd expect to be outright periodic. And periodicity is violated immediately with the dimensions changing.