roxics
Veteran
http://petapixel.com/2014/01/22/jpe...port-12-bit-color-depth-lossless-compression/
What could this possibly mean for mjpeg video?
What could this possibly mean for mjpeg video?
JPEG gets 12 bit update
- Thread starter roxics
- Start date
High1ander
Well-known member
Well, not just 12bit, but also lossless compression too...
I kinda like working with Jpegs knowing that they are compressed and lacking alpha channel, etc. Because then I know when people send me files (like a logo) or I work with them that I can say, "hey, this file is compressed and not RAW, etc." can you send me the original, like a Tiff, PNG or vector format. Now it will be harder for me to figure out if it is good or bad
I mean, its like when someone says they shoot something in HD, can be full HD, 720p, shot on a shitty cell phone, etc.
I kinda like working with Jpegs knowing that they are compressed and lacking alpha channel, etc. Because then I know when people send me files (like a logo) or I work with them that I can say, "hey, this file is compressed and not RAW, etc." can you send me the original, like a Tiff, PNG or vector format. Now it will be harder for me to figure out if it is good or bad
I mean, its like when someone says they shoot something in HD, can be full HD, 720p, shot on a shitty cell phone, etc.
G
Guest
Guest
I look forward to more bits. For both mjpeg and stills. dslr's like D800 really are pushing the 8bit jpeg.
there is already so many format for high-res, uncompressed picture.
Jpeg is supposed to be the high-compressed picture you can attch into email or fill a usb key with thousand of them.
So now we will have a mean to produce jpeg that are incompatibles with most TV, DVD player, camera, programs.... uselless for me.
Jpeg is supposed to be the high-compressed picture you can attch into email or fill a usb key with thousand of them.
So now we will have a mean to produce jpeg that are incompatibles with most TV, DVD player, camera, programs.... uselless for me.
combatentropy
Veteran
This update might bring about an ideal video format: something like CinemaDNG, in that it's a simple sequence of still images, but also something like ProRes, in that the images are mildly compressed. It would also be royalty free, coming from the JPEG group.
I've given the subject years of thought and decided that a format that meets these specs would be ideal:
10 bit. 8 bits is too few for grading. 12 bits is nice but 20% more space than 10 bit. And 10 bit encoded logarithmically would devote enough bits to each F-stop to allow enough gradability in post, as Apple ProRes has shown.
2K. 4K is nice but four times the price, and 2K is about what the human eye can see for the screens I'd be targeting. Also, you could upres it if need be to smoothe out the jaggies for those on the front row.
4:2:2. 4:4:4 is nice but 50% more space, and most people, indeed most pros, can't see a difference. 4:2:2 is great, again as ProRes has shown.
1 GB/min. This is a gift to the cameraman. No more math needed to know how much you can fit on a card. You can tell just by glancing at the label. 64 GB = 64 min, 128 GB = 128 min. This also allows recording on commodity cards like SDXC and CF, because the data rate would be about 17 MB/sec (133 Mbps). You could even record slow motion, 60-90 fps, on some of the faster cards. And at about $1 / GB, you can afford to just put the card on the shelf when you're done. The acquisition format is the archive format. Simplified workflow all around. (To reach 1 GB/min. the compression ratio for 10-bit, 2K, 4:2:2 video is only about 7.5:1, which should still be visually lossless, especially with a modern codec).
Computationally unintensive. Another advantage of a JPEG-based codec is that your computer won't overheat while working on it. Fast rendering, cheaper hardware, etc. (again, like ProRes)
I've given the subject years of thought and decided that a format that meets these specs would be ideal:
10 bit. 8 bits is too few for grading. 12 bits is nice but 20% more space than 10 bit. And 10 bit encoded logarithmically would devote enough bits to each F-stop to allow enough gradability in post, as Apple ProRes has shown.
2K. 4K is nice but four times the price, and 2K is about what the human eye can see for the screens I'd be targeting. Also, you could upres it if need be to smoothe out the jaggies for those on the front row.
4:2:2. 4:4:4 is nice but 50% more space, and most people, indeed most pros, can't see a difference. 4:2:2 is great, again as ProRes has shown.
1 GB/min. This is a gift to the cameraman. No more math needed to know how much you can fit on a card. You can tell just by glancing at the label. 64 GB = 64 min, 128 GB = 128 min. This also allows recording on commodity cards like SDXC and CF, because the data rate would be about 17 MB/sec (133 Mbps). You could even record slow motion, 60-90 fps, on some of the faster cards. And at about $1 / GB, you can afford to just put the card on the shelf when you're done. The acquisition format is the archive format. Simplified workflow all around. (To reach 1 GB/min. the compression ratio for 10-bit, 2K, 4:2:2 video is only about 7.5:1, which should still be visually lossless, especially with a modern codec).
Computationally unintensive. Another advantage of a JPEG-based codec is that your computer won't overheat while working on it. Fast rendering, cheaper hardware, etc. (again, like ProRes)
Last edited:
combatentropy
Veteran
Really? Can you point me to an article? Everything I've read says that wavelets are much more intense.
Samuel H
Major Contributor
I have programmed wavelet encoders. Really, it doesn't get much simpler than that. All it does is: calculate sum and difference between each pair of contiguous pixels, store that in a matrix; go through that matrix, calculating sum and difference of each two contiguous elements; go through that new matrix, calculating sum and difference of each two contiguous elements. Repeat a couple of times more. Throw away small numbers. Done. Much faster than the discreet cosine transform that jpeg uses.
Last edited: