Convert JPEG to RAW

[rpcrowe]

No that's not a typo error Topaz is claiming that A.L. in their program https://topazlabs.com/jpeg-to-raw-ai/ref/81/ will convert JPEG images to RAW This seems very unlikely to me...

Although, I frequently use the Camera RAW filter in Adobe Photoshop filter selection drop down menu when I am editing JPEG images. I really wonder how, even with A.I., the Topaz program can replace the information that was deleted in converting a image from RAW to JPEG or in the initial capture of an image when the photographer is shooting in JPEG!

If this is possible, it might be nice for sports shooters who initially shoot in JPEG to take advantage of the added imagery allowed by some camera's buffers or who need JPEG for quick transmission to editors at the headquarters and then want the added quality of editing a RAW image later in the work flow...

[tclune]

Try it. If you get the Brooklyn Bridge artifact, don't be surprised.

[xpatUSA]

No that's not a typo error Topaz is claiming that A.L. in their program https://topazlabs.com/jpeg-to-raw-ai/ref/81/ will convert JPEG images to RAW This seems very unlikely to me...

Easy enough if the conversion matrices and/or curves are known.
Indeed, for certain cameras not used by "most of us" the matrices, etc, are available in the proprietary raw files and could be applied back-asswards to approximate the raw data. Nothing I would want to do but most certainly possible ...

[DanK]

approximate

I think that is the key word. Data lost are data lost.

[george013]

No that's not a typo error Topaz is claiming that A.L. in their program https://topazlabs.com/jpeg-to-raw-ai/ref/81/ will convert JPEG images to RAW This seems very unlikely to me...

Although, I frequently use the Camera RAW filter in Adobe Photoshop filter selection drop down menu when I am editing JPEG images. I really wonder how, even with A.I., the Topaz program can replace the information that was deleted in converting a image from RAW to JPEG or in the initial capture of an image when the photographer is shooting in JPEG!

If this is possible, it might be nice for sports shooters who initially shoot in JPEG to take advantage of the added imagery allowed by some camera's buffers or who need JPEG for quick transmission to editors at the headquarters and then want the added quality of editing a RAW image later in the work flow...

A raw converter exist out of 2 parts: the converter to create a rgb raster image and an editing part to work on that raster image. If you load a jpg in a raw converter there will be some functions you can't use.

Not sure, but I think jpg can be converted to dng too. What has been said, try it.

George

[xpatUSA]

Not sure, but I think JPEG can be converted to DNG too.

My old-ish version of the Adobe Converter does not see JPEGs. In other words, it ignores them and does not even attempt to convert them.

[DanK]

To clarify my earlier remark: I have no idea whether a JPEG can be altered to match the file specs of a given raw file type. My comment was merely that the data lost in converting it to JPEG in the first place is gone, so it isn't clear to me what benefit there would be to converting it.

Both Lightroom and ACR (which are the same processing engines) allow one to use precisely the same editing tools on any file the software can read, including JPEGs and raw files. They just don't work as well with JPEGs.

[Manfred M]

It looks like what Topaz is likely doing is using their "AI" technology to take a JPEG (8-bit and lossy compression) and using it to emulate what the original raw data might have looked like.

At this point what I see is a solution offered by Topaz looking to solve a problem that may not exist. How well it does this will take some serious testing on the part of someone quite knowledgeable.

[xpatUSA]

It looks like what Topaz is likely doing is using their "AI" technology to take a JPEG (8-bit and lossy compression) and using it to emulate what the original raw data might have looked like.

At this point what I see is a solution offered by Topaz looking to solve a problem that may not exist. How well it does this will take some serious testing on the part of someone quite knowledgeable.

I'm not a Topaz user. I find it curious that the emphasis in this thread is on JPEG. Can Topaz also "un-convert" TIFF, PNG, GIF, et al?

[MrB]

I don't understand much of the theoretical stuff with regard to file types, so this is a question:
When a JPEG is opened as a raw file in PS Elements, it goes into Camera Raw and then emerges in Elements as a 16-bit file (and can be saved as a 16-bit TIFF); is there any advantage to editing this in Elements, compared with opening the original 8-bit JPEG image file directly in Elements for editing?

Philip

[Manfred M]

I don't understand much of the theoretical stuff with regard to file types, so this is a question:
When a JPEG is opened as a raw file in PS Elements, it goes into Camera Raw and then emerges in Elements as a 16-bit file (and can be saved as a 16-bit TIFF); is there any advantage to editing this in Elements, compared with opening the original 8-bit JPEG image file directly in Elements for editing?

Philip

There is a theoretical advantage when moving to 16-bit as rounding errors will be smaller than working in 8-bit. BUT all that is really happening is that your 8-bit value JPEG is being translated into a 16-bit value. The values will be identical, but just stored at a higher precision.

11111111 -> 0000000011111111

[pnodrog]

There is a theoretical advantage when moving to 16-bit as rounding errors will be smaller than working in 8-bit. BUT all that is really happening is that your 8-bit value JPEG is being translated into a 16-bit value. The values will be identical, but just stored at a higher precision.

11111111 -> 0000000011111111

Be a "bit" careful.

The msb (most significant bit) in image files is the mid tone value regardless of whether a 8,12 or 16 bit range is in use. Intel place the msb at the right-hand end and the rest of the world places it at the left-hand end to align with all our other numbering systems. The lsb (least significant bit) will be the smallest tonal value available. (probably mainly noise)

So for the rest of the world scaling from 8 to 16 bits would be: 11111111 -> 1111111100000000 The added bits being the lesser values.

[xpatUSA]

Be a "bit" careful.

The msb (most significant bit) in image files is the mid tone value regardless of whether a 8,12 or 16 bit ranges is in use. Intel place the msb at the right-hand end and the rest of the world places it at the left-hand end to align with all our other numbering systems. The lsb (least significant bit) will be the smallest tonal value available. (probably mainly noise)

So for the rest of the world scaling from 8 to 16 bits would be: 11111111 -> 1111111100000000 The added bits being the lesser values.

Good points. Also not forgetting that some systems have a sign bit at one end - didn't Adobe do that sometime?

I used to program the BBC model 'B' in machine code just for fun - and I remember 16-bit integer arithmetic vaguely but just can't remember which byte came first in the world of the 6502 processor ... anybody?

Then along came the Mac with it's Motorola 68000 and that processor was too much for me. Typically, I changed to Forth '83 while everybody else changed to 'C' and objects. Another boat missed ...

[george013]

The msb (most significant bit) in image files is the mid tone value regardless of whether a 8,12 or 16 bit ranges is in use.

That would mean that the mid tone value is always a 1 or a 0?

This info is only important for the programmer who's writing his code. And even then when he's writing his code at bit level.

The difference in image quality when using a bit depth more than 8 is becoming important when doing a lot of editing. Or when you've an output device that deals with more than 8 bit.

I stopped working in Pascal when Adobe wrote it's first ps in Pascal. They later switched to something else.

George

[pnodrog]

Actually calling the msb the mid tone is a very risky simplification. Safer for me to call it the mid value. Any tonal adjustment (exposure, curves, levels, white point etc) in PP can change the value that is being used as the mid tone. However for a properly exposed image it is approximately where the mid tone will be.

[george013]

Actually calling the msb the mid tone is a very risky simplification. Safer for me to call it the mid value. Any tonal adjustment (exposure, curves, levels, white point etc) in PP can change the value that is being used as the mid tone. However for a properly exposed image it is approximately where the mid tone will be.

I know what you meant. But the value of the msb is 1 or 0. What you mean is that in a numerical variable the mid tone or value is the msb followed by zero's. That will give 128 in a 8 bit variable.

That's knowledge important for the programmer working on bit level. Not for people who want to convert an image from a to b or change the bit depth from x to y.

If you like it, here's the source code of photoshop 1.0.1 https://github.com/amix/photoshop

George

[xpatUSA]

Lest we all forget, assuming that "the mid tone" means mid-gray, it's value depends on the camera's ISO-type, aka EXIF tag 0x8830 'SensitivityType' and that value varies:

Saturation-based: about 99/255
Standard Output Sensitivity (SOS): about 118/255
Recommended Exposure Index (REI): whatever the camera manufacture decides is best after some testing.

In other words, 128/255d (10000000b) is not 'mid-gray', other than by co-incidence ...

99/255 might raise eyebrows, but ISO includes half a stop headroom for saturation-based; their SOS does not.