4 Replies Latest reply on Feb 10, 2010 4:52 PM by Detail Man

    Does DNG Converter 5.6 Alter WB Info of DMC-LX3 RW2s?

    Detail Man

      I have converted a Panasonic DMC-LX3 generated ".RW2" to DNG using DNG Converter 5.6 for the exclusive purpose of being able to open it in the current version (2.1040) of Rawnalyze. The conversion prodeeds OK, and I am able to view the resulting DNG in Rawnalyze. It is my (humble, and barely-informed) understanding that the DNG Converter 5.6 performs de-mosaicing upon the RW2. However, I am uncertain as to:

       

      (1) Whether or not the DNG Converter 5.6 has altered any White Balance information contained in the original DMC-LX3 RW2 image-file; and

       

      (2) Whether any such alterations as described above in (1) result in an alteration of the "Shot WB Coefficients" (that I am assuming represent the DMC-LXW "raw" mode White Balance Scales) displayed when Rawnalyze is in Histogram display mode.


      My goal is to accomplish a "UniWB" scheme using the DMC-LX3 - by setting the "Custom White Balance" in the DMC-LX3 to a self-created color-target image displayed on my computer's LCD/TFT flat-screen display. See:

      http://ricohforum.com/phpbb/viewtopic.php?f=22&t=1556&start=50#p17852

       

      My original self-created color-target image was derived from the DxO Mark database for the DMC-LX3 Firmware Level 1.0 first published on or around November 17, 2008. See:

      http://ricohforum.com/phpbb/viewtopic.php?f=22&t=1556&start=40#p17389

      for my initial details (including a link to the DxO Mark database entry for the DMC-LX3).

       

      The DxO Mark White Balance Coefficients are: Red = 2.55; Green = 1.00; and Blue = 1.52.

       

      When I used the relative weightings listed directly above to create a color-target (using the DMC-LX3 at the latest Firmware Level 2.1, and using the "Standard Color" selection within the "Film Mode" settings: Contrast = 0; Saturation = 0; Sharpness = -2; and Noise Reduction = -2) ...

       

      the resulting DNG (converted from a DMC-LX3 RW2) shows "Shot WB Coefficients" as: Red = 1.0760; Green = 1.0000, Green = 1.0000; and Blue =1.1597 when viewed in Rawnalyze Histogram display mode.

       

      My first (humble, and barely-informed) assumption from the above result is that (perhaps) I need to re-scale the DxO Mark White Balance Coefficients used in my initial experiment by a factor equal to what Rawnalyze reports as the "Shot WB Coefficients". When I do such a re-scaling (shown here rounded to four digits following the decimal point - despite the fact that the original DxO Mark data is only precise to two digits following the decimal point), I get:

       

      Red = 2.370


      Green = 1.0000


      Blue = 1.3107

       

      I am thinking that I should make a new color-target for setting the DMC-LX3 Custom White Balance where the relative weighting of Red, Green, and Blue in the color-target reflects the above-listed modified RGB Coefficients. In fact, I did just that (as you will find if you read the first Ricoh Forum post that I cite above) ...

       

      However, I am (just barely) wise enough to recognize how little I know about these things (details surrounding "raw" image-files in general, the Adobe DNG Converter 5.6, and the very interesting Rawnalyze 2.1040. I want my action sand information disseminated to be accurate (for the benefit of my own DMC-LX3 "UniWB" implementation scheme, as well for the benefit of any readers who might assume that I actually may know something about what I am posting!).

       

      Gabor, I know that you are out there somewhere (it appears only a few hundred miles to the north of me in Seattle). I have been doing all the "Googling" and forum-reading that I can to try to understand the details of these matters (including reading your Rawnalyze on-line help information). If you have the time - would you be so kind to enlighten me as to what your understanding of these matters is (or, if possible, refer me to information sources where I might be able to be able to figure these specific questions out on my own). Many thanks!

        • 1. Re: Does DNG Converter 5.6 Alter WB Info of DMC-LX3 RW2s?
          Panoholic Level 2
          function(){return A.apply(null,[this].concat($A(arguments)))}

          It is my (humble, and barely-informed) understanding that the DNG Converter 5.6 performs de-mosaicing upon the RW2


          You could convert it in demosaiced format (it is an option under Preferences, if you select the compatibility as "Custom"); however, if Rawnalyze can display the DNG, then it is certainly not demosaiced.

           

          function(){return A.apply(null,[this].concat($A(arguments)))}

          Whether or not the DNG Converter 5.6 has altered any White Balance information contained in the original DMC-LX3 RW2 image-file


          The converter usually changes the temperature and tint; however, Rawnalyze displays the coefficients of the raw channels, which are stored in the raw file either by the camera or by the raw/DNG converter.

           

          function(){return A.apply(null,[this].concat($A(arguments)))}

          the resulting DNG (converted from a DMC-LX3 RW2) shows "Shot WB Coefficients" as: Red = 1.0760; Green = 1.0000, Green = 1.0000; and Blue =1.1597 when viewed in Rawnalyze Histogram display mode


          This is not excellent, but not horrendeous either. Have you tried using a totally clipped raw image as white balance template? It works with some cameras, but not with all. Give it a try and post the results pls:

           

          a. shoot a white sheet with +4 EV bias,

           

          b. verify in Rawnalyze if everything was clipped,

           

          c. load that raw image in the camera as WB template,

           

          d. make a shot with "custom WB" or whatever the name of that option is with your camera, i.e. using the previous shot as WB template,

           

          e. verify the WB coefficients; they should be very close to 1.

           

          Gabor

          • 2. Re: Does DNG Converter 5.6 Alter WB Info of DMC-LX3 RW2s?
            Detail Man Level 1

            Gabor,

             

            The problem with your suggested procedure is that (unlike Nikons), on the Panasonic DMC-LX3 image-files cannot be placed on the camera's memory card and used as "templates" (as the Step 'c' in your suggested procedure requires):


            "c. load that raw image in the camera as WB template"

             

            It is for this reason that it appears (to me) that with the DMC-LX3 my only option is to generate a "Custom White Balance" by aiming the camera at particular pre-determined color-balanced target (such as a cutom-made color target displayed on my computer's LCD display)

             

            Thus, my question revolves around the origin and meaning of the "Shot WB Coefficients" in your Rawnalyzer Histogram display. When you say:

             

            "Rawnalyze displays the coefficients of the raw channels, which are stored in the raw file either by the camera or by the raw/DNG converter."

             

            in your reply, are you saying that Rawnalyzer's displayed "Shot WB Coefficients":

             

            (1) Reflect numerical coefficient values that (only) the camera itself generates; or

             

            (2) Reflect numerical coefficient values that the DNG Converter 5.6 definitely alters to be different from the values of the coefficients that the camera reports; or

             

            (3) Reflect numerical coefficient values that the DNG Converter 5.6 may alter to be different from the values of the coefficients that the camera reports ?

             

             

            Please pardon my profound lack of knowledge about these things, but - it is not clear to me when you state:

             

            "The converter usually changes the temperature and tint"

             

            whether you (may) have intended to mean:

             

            (1) That the values of the "Shot WB Coefficients" themselves are altered by such changes in "the temperature and tint"; or

             

            (2) Whether the values of the "Shot WB Coefficients" themselves are independent from such changes in "the temperature and tint".

             

            Thank you very much for your time and thoughts!

            • 3. Re: Does DNG Converter 5.6 Alter WB Info of DMC-LX3 RW2s?
              Panoholic Level 2

              This issue is a problem for me, for I don't have the camera, nor the manual. I made a search in Internet and found in one review

               

              White balance can be set by color temperature or a white/gray card

               

              This means, that you can use an existing shot as white balance basis; this is a very normal way of WBing shots, particularly when creating in-camera JPEG: you can make a shot of a white sheet and declare that as WB template. (My understanding is, that you can store even two independent images as WB templates and select from them.)

               

              If the problem is, that you can not copy a raw file on the camera's card, then you have to keep the WB template there (I guess you can make it "protected", to prevent it from being deleted; I did that too). I guess the camera won't copy any image on the card, but you can do that in your operating system.

               

              Re the WB coefficients: as Rawnalyze does not support the native raw files of your camera, you need to convert them in DNG. The DNG converter does convert the white balance values in coefficients; if you load such a file in Rawnalyze and go to the Histogram panel, you can see the coefficients of the raw channels.

               

              The coefficients are the effective white balance values; white balancing occurs by multiplying the pixel values by a channel related constant. Not the coefficients themselves but the values relative to each other determine the white. For example (red=2, green=1, blue=1.5) and (red=1, green=0.5, blue=0.75) are equivalent in this sense (the multiplication may cause "clipping", but that's another issue).

               

              The relative coefficients are independent (or they are supposed to be independent) of the data format, i.e. native raw or DNG, unlike the temperature and tint.

               

              Gabor

              • 4. Re: Does DNG Converter 5.6 Alter WB Info of DMC-LX3 RW2s?
                Detail Man Level 1

                Gabor,

                 

                Thanks for the information!

                 

                "The relative coefficients are independent (or they are supposed to be independent) of the data format, i.e. native raw or DNG, unlike the temperature and tint."

                 

                This statement of yours above answers my question. It implies that (assuming the display coefficients are, in fact, independent, as you speculate is likely the case, above), I can proceed by further adjusting the relative weighting of the colors on the color-target images (that I create in order to display on my LCD display, and then use that adjusted color-target image to set one of the two available "Custom WB" settings on the DMC-LX3). I will proceed (as I already tentatively have done so) by applying the (Rawnalyzer indicated) required correction factors upon the original relative color weightings of the "color-target" image (that I displayed on my LCD display and then "sampled" with the LX3 "set Custom WB" procedure).

                 

                I recognize that my camera's chosen user-settings for Contrast, Sharpness, Noise Reduction, and perhaps Saturation may have a minor effect on the on-camera "Live Histogram" display as well. However, it appears from reading your and Guillermo's discussions on the Luminous Landscape discussion thread beginning at:

                http://luminous-landscape.com/forum/index.php?showtopic=22250&st=0

                (as well as some information from other sources) that the above listed camera settings have a minor effect relative to that of the "WB" coefficients for RED, GREEN, AND, BLUE. It may be that my particular favorite settings (for the purpose of shooting in-camera JPGs) have contributed to the discrepancies noted.

                 

                Since your Rawnalyzer's "Shot WB Coefficients" (resulting from looking at the DNG converted DMC-LX3 RW2 image-file of a real-world recorded image) indicate the RED coefficient at 1.0760 and the BLUE coefficient at 1.1597, I have adjusted the DxO Mark database coefficients from the "Color Response" Tab and Section at the web-page:

                http://www.dxomark.com/index.php/eng/Image-Quality-Database/Panasonic/Lumix-DMC-LX3#

                from their originally published values - reducing the value of the RED and BLUE relative color-weightings (contained in my created color-target image for display on my LCD screen) by a factor of 1.0760 and 1.1597, respectively.

                 

                Regarding the DMC-LX3 (and other Panasonic cameras such as the LZ5, FZ30, FZ50, and TZ4 that I own or have owned). The procedure for setting "Custom WB" involves the camera storing a special sample of the "target" that appears to the user as a small rectangular area contained within the larger area of the normal image-frame when the user performs the "set custom WB" procedure. Individual Image-files cannot be used for the "WB template".

                 

                The sampled information (that the cameras collect in the above mentioned procedure) are stored within internal memory of the cameras (and are not stored on removable memory cards, or the small built-in on-board memory space). Thus, it is "internally held" information that is not available as a stand-alone image-file that can be created/copied/moved/deleted, etc.

                 

                After posting information regarding my "UniWB" education and attempts to implement it using the DMC-LX3 at:

                http://ricohforum.com/phpbb/viewtopic.php?f=22&t=1556&start=40#p17389

                and later at:

                http://ricohforum.com/phpbb/viewtopic.php?f=22&t=1556&start=50#p1785

                I may well be updating my findings (in theory, and once effectively reduced to actual practice).

                 

                (It presently appears) that the revised "color-target" JPG image that I posted at the latter link above does correctly implement the re-scaling of the relative weights of the RGB WB coefficients that appear (thanks to your kind assistance) likely to be more appropriate for my DMC-LX3 (at Firmware Level 2.1, and using my particular camera settings listed above.)

                 

                More actual practice with further test shots will reveal whether further iteration(s) are in order to converge within some small error-band of the ideal ("Shot WB Coefficients"), being - RED = 1.0000, GREENs = 1.0000, BLUE = 1.0000.

                 

                It's good to know that the DNG Converter 5.6 does not (as you said) perform de-mosaicing (other information sources have claimed that such is the case). I presume that is why Rawnalyzer's two separate sets of displayed WB coefficient values reflect the (pre-de-mosaiced) state of the data (with two separate GREEN coefficient values, and one RED and BLUE value). I presume that this (may, in some way, possibly) reflect the physical layout of the image-sensor (where there exists two GREEN photo-sensors for each RED and BLUE photo-sensor). Then again, maybe my speculation is wrong about why there are two separate GREEN WB coefficients reported by the Rawnalyzer Histogram window data.


                I really appreciate your thoughts and assistance regarding your knowledge about these things. Thanks again!