Curiosity only: If we could print a RAW file, would it look better then a printed JPEG file?

[Mike_PEAT]

Load the RAW file into an editing program, and select PRINT! It's not a JPEG, it's not a TIFF...it's just the RAW data and the editing program turning that data into a visible image.

Don't know what's with all these people saying you can't print a RAW file!

If you save it to a format like PSD in Photoshop you don't lose anything.

 

[John Deerfield]

Mike_PEAT wrote:

Load the RAW file into an editing program, and select PRINT! It's not a JPEG, it's not a TIFF...it's just the RAW data and the editing program turning that data into a visible image.

At which point the the editing program converted the file to an 8-bit image file. Period. And since it's an 8-bit file, color information is being thrown away. Period. If you are on a Mac with an Epson printer and the 16-bit driver, you could print in 16-bit. If you use a professional lab that prints in 16-bit, you can get a 16-bit print but that lab still wouldn't accept a Raw file because the Raw file is NOT an image format.

Don't know what's with all these people saying you can't print a RAW file!

Because you can't. That's why.

If you save it to a format like PSD in Photoshop you don't lose anything.

Provided you saved it as a 16-bit file! If you saved it as an 8-bit file, you are throwing out data.

 

[Sandy]

JPEG throws away millions of colours? Doesn't that depend on the colour space selected for creation of the JPEG? A JPEG will result in some degree of data loss - though this can be mitigated by minimal compression.

 

[John Deerfield]

Sandy wrote:

JPEG throws away millions of colours?

Yes.

Doesn't that depend on the colour space selected for creation of the JPEG?

No.

A JPEG will result in some degree of data loss - though this can be mitigated by minimal compression.

A JPEG will result in trillions of colors lost. And this can't be minimized via compression. Compression happens after the colors are already lost and performs even more color loss. But here's the thing: 99.99% of the time you can't see this. It is estimated that the human eye can see something like 10-million colors. An 8-bit JPEG is capable of over 16-million colors. Already more than we humans can distinguish. A 16-bit file is capable of trillions of colors. So, converting a 16-bit file to an 8-bit file results in the loss of a lot of colors! Yet an 8-bit file is still more than we can see. All these extra bits matter when you are do a lot of color grading/editing to an image. If I edit a blue sky with an 8-bit image, I might quickly start to see banding in the colors of the sky. If I edit in 16-bits, with it's trillions of more colors available, I can edit without getting the banding. The color space has nothing to do with it (until output I suppose!).

 

[ARShutterbug]

When I click on the icon, the software loads! I don't know what all these people are talking about when they say that they need to compile code!

 

[apaflo]

Mike_PEAT wrote:

Load the RAW file into an editing program, and select PRINT! It's not a JPEG, it's not a TIFF...it's just the RAW data and the editing program turning that data into a visible image.

Don't know what's with all these people saying you can't print a RAW file!

If you save it to a format like PSD in Photoshop you don't lose anything.

John Deerfield has posted some very good discussion on this, and in particular this statement: "A JPEG will result in trillions of colors lost."

The problem is that folks may not realize how that actually can be true, and don't accept it as fact. Let me walk through some numbers that show why John says that.

A 14-bit RAW file encoding color using the Bayer Color Filter method is "demosaiced" to produce an RGB image file of the same number of pixels as the sensor contained "sensels" (light sensitive locations that were recorded). But there is NOT a direct relationship between the color recorded for a given sensel and the pixel color or luminance value for a pixel with the same coordinates in the resulting image!

The processing (demosaicing) requires a minimum of a 3x3 matrix of 14-bit data values to produce one 16-bit three channel (RGB) pixel value. The 3x3 14-bit matrix has 126 bits, the 16-bit pixel has 48 bits. That is, in the RAW file there is data that could produce a total of 8.507e+37 colors, but that is trimmed to only 2.815e+14 possible colors for the RGB image data. So the loss is 1 times 10 with 22 zeros tacked onto it. (Almost a trillion trillions! And if a 4x4 matrix is used rather than 3x3, it is more than a trillion trillion.)

Note that if the file is then converted to an 8 bit format such as JPEG, it is reduced to only three 8-bit channels (rather than the16-bit channels calculated above, which would apply to PSD or TIFF formats), and that is 1.678e+7 (the commonly heard 16.7 million color values for JPEG). And whether a printer can use a 16-bit input file or not, it uses only 8-bit values to print the image, so that is as good as it gets.

The issue with a printer and the number of colors is not how many, because we are not likely to be able to distinquish between more than the 16.7 million the printer can produce. It's a matter of which colors, and each step of the process from the trillions and trillions available in the raw data down to the few we actually will see, requires that decisions be made on what is saved and what is not saved for the next step in the process.

 

[hedwards]

Nope, I've seen it in cases where the JPG is created at the end and not a moment before. It's something that happens rarely, but looks horribly when it happens.

The times where you'll see it rear it's ugly head are where you have color gradients changing quickly enough that the sampling results in a stair step in two directions. You end up with these ugly dots. I'll have to keep looking for the image, but the times I've noticed it, I'm positive I wasn't recompressing an image as I never do that.


That being said, you're definitely right about not wanting to work with compressed files if you can avoid doing so, but that is a different type of degradation that happens.

It's been a while since I've seen it, hopefully I can locate the image I'm thinking of, but really this is an academic exercise as you're not going to be printing large enough that you're seeing the JPG sample plots normally anyways. With 12mp sensors and larger, most folks aren't going to need to print the largest size possible. And if you can't see them, then you're not going to have issues with it.

 

[apaflo]

hedwards wrote:

Nope, I've seen it in cases where the JPG is created at the end and not a moment before. It's something that happens rarely, but looks horribly when it happens.

But it is still just as I suggested, and extra "edit" because the image is resampled for display. That can be on a computer monitor or by a printer.

The times where you'll see it rear it's ugly head are where you have color gradients changing quickly enough that the sampling results in a stair step in two directions. You end up with these ugly dots. I'll have to keep looking for the image, but the times I've noticed it, I'm positive I wasn't recompressing an image as I never do that.

That being said, you're definitely right about not wanting to work with compressed files if you can avoid doing so, but that is a different type of degradation that happens.

It's been a while since I've seen it, hopefully I can locate the image I'm thinking of, but really this is an academic exercise as you're not going to be printing large enough that you're seeing the JPG sample plots normally anyways. With 12mp sensors and larger, most folks aren't going to need to print the largest size possible. And if you can't see them, then you're not going to have issues with it.

 

[Roberto de La Tour]

Mike_PEAT wrote:

Roberto de La Tour wrote:

IN practice (what my eyes can see): perfectly OK to print from JPEG. No visible loss. Shoot in RAW, post process from this raw if needed, look at it well, convert to JPEG maximum res/quality, look again, and print. It will be just fine even using a very good printer

If you print the 12-14 bit RAW file on a top of the line printer, compared to printing an 8 bit JPEG, and you WILL see a difference if the RAW file contains colours the JPEG doesn't.

When you save an image as a JPEG millions of colours are thrown away!

Hello

Maybe... All I say is that the difference is not really visible on my monitor (Macbook Pro 15"). OK, I admit I have not a Retina screen. And once printed (JPEG maximum) I do not perceive a loss of colours

Im do not see a difference if I save in TIFF either


Cheers
Roberto