How to determine what color temperature was used to develop an image?

[aajax]

The subject question pretty directly raises another much more basic question which is "what reference information does Canon provide regarding the metadata produced by Canon cameras and software (e.g., DPP)?".

A basic experiment that I've undertaken is to open a raw image file in DPP4 and then adjust the White Balance by changing the "Color Temperature".  Then the raw file is "Converted and saved" (some might call that developed) in a standard image format (e.g., in this case 16bit .tif).  When I examine the metadata (using ExifTool/ExifToolGUI) I cannot find anything in the .tif file that indicates what "Color Temperature" was used by DPP4 to develop said image.  Isn't this something that should be possible?

The metadata item referred to as "Color Temperature" says 5200 in both the raw and .tif file.  Hasn't this been changed?

From some research I've done on White Balance I think it is accurate to say that "Color Temperature" is NOT a property of the raw data.  Rather, the process of developing the raw data into an image (file) involves the application of color temperature.  Of course color temperature is a property that the camera must assign in order to develop the raw data and create an image (i.e., .jpg) file.  I have the idea that this might be a lot more complicated that anything I've yet learned but it does seem like I should be able to determine what value was used for that property.

It is also my understanding that DPP4 attempts/intends to mimic what the camera would do as the starting point for transforming the raw data into something that can be viewed as an image on the computer's display.  In my case, the White Balance tool always defaults to something called "Auto: Ambience Priority".  Whenever, this selection is changed to "Color Temperature" the value is set to 5200 (i.e., same as the metadata).  However, this selection does normally produce some change in the appearance of the image which suggests that 5200 was NOT the value used when "Auto: Ambience Priority" was selected.  I think what I'd like to be able to figure out is what temperature value corresponds to the image I'm looking at.  Right now I assume that if I switch to "Color Temperature" mode and select a value for temperature that would be it.  However, I haven't figured out how to tell what it is from the image file (e.g., .jpg or .tif) created by the development process.

It might be worth mentioning that other image editing software that I've tried suffers from the same problem!

[Waddizzle]

What does your application say about the WB of this JPEG file?

I am really curious.  LIke I said just above, the application must make a set of assumptions, most of which are false.

[aajax]

---

@Waddizzle wrote:

What does your application say about the WB of this JPEG file?

...

I am really curious.  LIke I said just above, the application must make a set of assumptions, most of which are false.

---

The answer is hard to come by because I haven't been able to figure out how to obtain a file from this website.  However, I did undertake a simple experiement which is to copy and paste a screenshot of the window in which ExifToolGUI is running to MS Paint and then save it as a jpg.  All of the metadata for that file follows:

---- ExifTool ----
ExifTool Version Number         : 10.39
---- File ----
File Name                       : MetadataExperiment.jpg
Directory                       : .
File Size                       : 269 kB
File Modification Date/Time     : 2018:08:21 16:57:03-04:00
File Access Date/Time           : 2018:08:21 16:57:03-04:00
File Creation Date/Time         : 2018:08:21 16:57:03-04:00
File Permissions                : rw-rw-rw-
File Type                       : JPEG
File Type Extension             : jpg
MIME Type                       : image/jpeg
Image Width                     : 1493
Image Height                    : 661
Encoding Process                : Baseline DCT, Huffman coding
Bits Per Sample                 : 8
Color Components                : 3
Y Cb Cr Sub Sampling            : YCbCr4:2:0 (2 2)
---- JFIF ----
JFIF Version                    : 1.01
Resolution Unit                 : inches
X Resolution                    : 96
Y Resolution                    : 96
---- Composite ----
Image Size                      : 1493x661
Megapixels                      : 0.987

I think this might be similar to the example you propose which makes the point that a jpg may have a very minimal amount of metadata and that metadata may include nothing about WB.  To my, very inexperienced, way of thinking WB is about how a camera converts the raw sensor data into an image which can be viewed and stored as a jpg file.  From the little bit of reading I've done, the concept of WB is also required to convert raw sensor data into a viewable image.

My reading includes this article which I think also makes your point about the cameras (also software attempting to automatically process raw data) ability, or maybe it would be better to say "inability", to precisely determine the WB.  Is this not the reason for those of us trying to improve on what the camera can do by itself to want to adjust the WB?

The more I've learned about digital photography and image editing the more amazed I am about how GOOD a job modern cameras do at producing a picture but still there is a little fun to be had by doing some added customization.

[ebiggs1]

A couple points that may have a bearing on what you want to know or perhaps not but here they are.

When editing an existing image, do not edit a JPG and then save it as a JPG even to do a simple crop.  First save the JPG as a TIFF and work on the TIFF.  It's really not going to kill the image saving it once. However, more often than not it can get re-edited numerous times, resized, cropped, etc. This path will in the end lead to image degradation. Working from a TIFF image format eliminates that. You can ruin images you value without understanding why.

Second contrary to popular belief you can not correct for every color. Every color effects every other color.  The ones that are opposite each other the least. The ones adjacent to each other the most. When equal amount of the three main colors are combined you get neutral grey.

How do we apply this?  It doesn't mater what the exif data reads as long as you can find neutral grey in the picture.  This puts equal portions of each color in the photo.  However, remember the first thing I said? A jpg may not have enough info left in its file to accomplish the desired results.  Especially if it has been opened and saved a few times. When you use Raw or TIFF you avoid that part.

Like I said this may or may not be the answer yuo seek.

[ebiggs1]

"First save the JPG as a TIFF and work on the TIFF."

This is the main reason to not use jpg.  Always shoot Raw. Raw maintains all its data and it is never altered,

[aajax]

Yes!  I do understand the difference between lossy and lossless compression.  For those who may not "lossy", which applies to JPG, means that when the data is uncompressed it is NOT restored to the same value as what was originally compressed.

Now that I have cameras capable of providing raw files that is all I use and my finished photos get kept in uncompressed TIFF format files.  However, as with most people I do have a sizable inventory of files, from my many past years and from friends, that originate in JPG format and now that I'm learning about digital image editing I have been following the rules you've prescribed for first converting to TIFF and then editing the TIFF files.

My original question was about how to figure out what color temperature was used to develop the raw sensor data into an image file of whatever kind.  Since cameras do develop raw sensor data and do produce JPG files this suggests that the answer to my question would involve understanding how cameras produce JPG.  I'm pretty sure that the same metadata that cameras put into the files, both raw and JPG, that they create can be preserved in TIFF files.  Isn't this the main reason photographers prefer TIFF to PNG format, which are both lossless formats?

As best I can now tell cameras generally, and for sure the ones I have, DO NOT preserve that value for color temperature in the form of metadata.  While the actual computational work involved in adjusting White Balance is still beyond my level of understanding it sure seems like it is within the realm of possibility to do this.  If so, it may be that until recently there was no known use for preserving such information.  Likewise, given all of the possible scenarios for post processing I can see that it would be near hopeless for such metadata to be accurately maintained throughout all possible subsequent editing steps.  However, this is also true for other metadata that is recorded.  Given the enormous amount of metadata that is being supplied by cameras, most of which is not yet useful to me, it remains a bit of a mystery to me as to how this one escaped such attention.

Because I'm developing the raw files using software I could undertake to preserve this information myself for those files.  However, I'm not experienced enough to know whether or not such an effort is warranted and of course this is no help for all those files that I did not develop.

[ebiggs1]

aajax,

You need to understand this as I repeat it. "It doesn't mater what the exif data reads as long as you can find neutral grey in the picture.  This puts equal portions of each color in the photo.  However, remember the first thing I said? A jpg may not have enough info left in its file to accomplish the desired results."

Being that you and I for than matter do hvae large inventories of jpg files, this is an issue.  99.9% of the time I shoot Raw and not jpg anymore. You probably do too. That is now and that was then.  Still when you do your edits you have to work with what you have and not what you wished you had. Forget the metadata file and do you edits according to my suggestion about finding as close of neutral grey area in each jpg. When I shoot today in Raw I really don't care where the WB is set.  I think it is on AWB.  I edit to neutral grey.  Almost all images have something that is or is close to neutral grey.

As to the metadata file incliding dpi, that is false. There is no dpi in a camera's image. All cameras shoot at max resolution.  Images are measured in absolute pixels. The dpi number is just a lable. it is not an indicator of actual pixels per inch. Dpi is meaningless to a computer monitor. Your monitor uses the number of pixels in the image recorded by the sensor.

[Waddizzle]

---

@aajax wrote:

---

@Waddizzle wrote:

What does your application say about the WB of this JPEG file?

...

I am really curious.  LIke I said just above, the application must make a set of assumptions, most of which are false.

---

The answer is hard to come by because I haven't been able to figure out how to obtain a file from this website.  However, I did undertake a simple experiement which is to copy and paste a screenshot of the window in which ExifToolGUI is running to MS Paint and then save it as a jpg.  All of the metadata for that file follows:

---

A basic JPG file is one type of image file,  Its’ metadata describes how the image is intended to be displayed.  The metadata contains no information about the substantive content of the image.

Your metadata contains DPI resolution data.  Altering these values will change the apparent size of the image.  All of the values in your metadata describe how the image should be displayed.  Not one is associated with image content.

[aajax]

---

@Waddizzle wrote:

What does your application say about the WB of this JPEG file?

...

---

While I couldn't figure out how to obtain the subject file from this website, it dawned on me that I could obtain the image.  As with the example I previously mentioned the metadata (created by MS Paint) says nothing about WB.  However, I thought it might be worth pointing out that I can still edit the WB and that presents the problem that caused me to post my original question.

For example, GIMP assumes a Color Temperature of 6500K and allows the user to update that with a known value if such exists.  That is what I haven't been able to figure out how to find.  However, using the GIMP provided Color Temperature tool, I can say I'd like to change it from 6500K to 4000K and that will produce the image that follows:

I can also use the same tool to specify that I want to change the Color Temperature from 9000K to 6500K (i.e., the same 2500K reduction).  The result, within the ability of my eyeballs to determine, is the same.  The implication of wanting the user to input 2 values is that it matters.  As I've said I don't know the particulars of the algorithm (i.e., computations) so I cannot explain why it matters which I suppose turns out to be the essence of my post.

By the way, I also know that Rawtherapee, which is another program that allows WB adjustment on already developed images (i.e., TIFF, JPG, PNG, etc.), only allows the user to specify a single input which is in units of degrees Kelvin.  However, in that case they assume the original temperature to be 6490K (i.e., pretty close to GIMP's default) with NO ability to change it.  This suggests that they're algorithm only cares about difference.

In conclusion, it looks to me like once developed the color temperature is lost and the only thing that can be done then is adjust it up or down but this will never allow an actual temperature value to be known.

[ebiggs1]

Perhaps I have misunderstood your entire question. You don't want to know how to set the correct WB, you want to know the actual Kevin temperature?  This I never tried to ascertain or wanted to know so I can't help.  Setting the "correct" WB is easy as I have explained.

[aajax]

Yes my question is about color temperature.  I am aware that there are other methods that the various image editing software products provide for setting/changing WB. Likewise, I don't have an opinion about which methods might be better than others which seems likely to depend on circumstances that may vary.  I'm just trying to understand the subject.

Based on my own study to date I've come to think that when the raw data is originally developed, whether by the camera when producing a JPG file or by myself using image editing software, that there is a color temperature that can be associated with the resulting image no matter what method is used to assign it.  In that, when you use the so-called "Click WB" method to tell the software where in the image a particular color (i.e., white, black, or grey) exists this allows the software to figure out what temperature corresponds to the RGB values for those pixels.  Insofar, as I do NOT know what those algorithms do it is very possible that this deduction of mine is what is mistaken.  That would certainly explain the idea that such a temperature is NOT actually known in all situations and therefore cannot always be provided.  If it is known, especially if in all situations, why wouldn't it be provided?  Such is the purpose of metadata.