12 Replies Latest reply: Apr 21, 2012 11:02 AM by Chris Cox RSS

    "wire removal" on a 16Bit (Half float) EXR file format

    prashanthvs Community Member

      HI. This query is with regard to a Wire-removal/Touchup process in dealing with16Bit(half float) EXR files. Here is the process we adopted.

       

      - Opened a 16Bit (Half float) Exr file in Photoshop, which showed it as 32Bit file

      - Since the footage has lot of fire/blast elements among which we need to erase a "WIRE", we need to use "Healing Brush" which doesn't work on 32Bit files. When we tried to 'Convert' 32 Bit file to 16Bit file directly, we get a 'Washed Out' problem, since the shot is over exposed due to Fire/Blast elements in it. So here is a pipeline we used.

       

      ---- Added "Exposure" (Image > Adjustment > Exposure)

      ---- Set the "Exposure" to -5 (since it is a over-exposed frame, value -5 depends on exposure values of shots)

      ---- File > Export > Render Video (OpenExr output)

      ---- Open the above render file in photoshop

      ---- Convert to 16bit (Image > mode > 16bit)

      ---- Use any 16 bit tool for cloning (Eg. Healing tool)

      ---- Added "Exposure" (Image > Adjustment > Exposure)

      ---- Set the "Exposure" to +5 (Reversing)

      ---- Convert to 32bit (Image > Mode > 32Bit)

      ---- File > Export > Render Video (OpenExr output)

       

      After rendering out, we checked the "original footage" with the "output exr" and apart from the Touch-up (wire removal) portion, all other details remain the same. Since this is a high-end hollywood project, can you guys (professionals) confirm whether this process is safe. Please let us know of any optimised workflow.

       

      Note1: During the process we found a bug that this process only works, when we render whole footage, meaning if we try single frame, then there seems to be a slight color difference only in the over exposed area. 

       

      Note2: When we opened the client input EXR file, under Nuke it shows as "16 Bit (Half float)" but the same when opened under photoshop it shows as 32 Bit file. Why is this?

        • 1. Re: "wire removal" on a 16Bit (Half float) EXR file format
          ibarrien Adobe Employee

          I think you've posted this in the wrong forum - where were you supposed to post it? As you've noticed there are many different Photoshop forums.

          If you let me know I can move it to the correct forum for you.

          • 2. Re: "wire removal" on a 16Bit (Half float) EXR file format
            prashanthvs Community Member

            It's with regard to Photoshop touch up (Wire removal) process. Kindly move this thread to concerned forum.

             

            Thanks,

            • 3. Re: "wire removal" on a 16Bit (Half float) EXR file format
              Community Member

              My one concern would be with automated healing and content aware filling.

               

              These automated techniques only consider the data in the present frame.  This can lead to apparent quirks when the frames are played back in succession. Nuke offers the ability to address multiple frames. Why not use Nuke if you have it available to you? You can avoid all those conversion steps by simply avoiding Photoshop for this task.

              • 4. Re: "wire removal" on a 16Bit (Half float) EXR file format
                prashanthvs Community Member

                Thanks for your reply.

                 

                While erasing the WIRE using "Healing" method, we carefully jump to before/after frames and check for any jitters. For your kind info, we understand exactly about the 'quirks' problem that may rise, but believe us, we do it very meticulously and also we match the Grains too. Basically we import the entire footage inside Photohop under 'Video' mode and play it over there itself.

                 

                No doubt that Nuke on the other hand is an amazing tool, but since the  "Quality Check" process in our vfx pipeline is very very higher (meaning, we have to achieve the at-most quality), we prefer Photoshop which is specifically meant to deal with image manipulation.

                 

                So, we request you to give out your answers on our main post.

                 


                • 5. Re: "wire removal" on a 16Bit (Half float) EXR file format
                  Community Member

                  How are you able to obtain the best quality if you are wrecking the color depth and toggling the exposure?

                   

                  Doesn't the clone tool work at your preferred bitdepth?

                  • 6. Re: "wire removal" on a 16Bit (Half float) EXR file format
                    Chris Cox Adobe Employee

                    First: you seem completely confused between bit depth in files and documents.

                    In Photoshop: 8 and 16 bit/channel are integer formats, while 32 bit/channel is a floating point format.

                    EXR stores data as 16, 24, or 32 bit/channel floating point - which Photoshop needs to open into it's 32 bit/channel mode.

                     

                    Converting from 32 bit/channel floating point to 16 bit/channel integer lost any overranged values, and may have applied toning to the image (making the process non-reversible).  Your exposure changes added a lot of quantization, so you probably have 7 bits or less of actual image data left.

                     

                    >> can you guys (professionals) confirm whether this process is safe.

                     

                    I can confirm that what you describe is NOT safe.

                     

                     

                    You really need to learn more about the formats and tools before trying something like that.

                    • 8. Re: "wire removal" on a 16Bit (Half float) EXR file format
                      prashanthvs Community Member

                      How are you able to obtain the best quality if you are wrecking the color depth and toggling the exposure?

                       

                      That's exactly my question is all about. But fortunatley, Photoshop seems to preserve the Clipped info even after changing the Bit depth. Here is an another example i tried.

                       

                      ---- Opened the EXR file in Photoshop

                      ---- Added "Exposure" (Image > Adjustment > Exposure)

                      ---- Set the "Exposure" to -5 (since it is a over-exposed frame, value -5 depends on exposure values of shots)

                      ---- File > Export > Render Video (OpenExr output)

                      ---- Open the above render file in photoshop

                      ---- Convert to 8bit (Image > mode > 8bit)

                      ---- Used Healing tool

                      ---- Added "Exposure" (Image > Adjustment > Exposure)

                      ---- Set the "Exposure" to +5 (Reversing)

                      ---- Convert to 32bit (Image > Mode > 32Bit)

                      ---- File > Export > Render Video (OpenExr output)

                       

                      Again the image doesn't seem to loose any quality. I guess Photoshop  might be following a "Concatenation" method which all Compositing softwares use.

                       

                      Doesn't the clone tool work at your preferred bitdepth?

                       

                      No. Certain tools in Photoshop are designed to work until 16Bit images.... yet they don't work in 32Bit image.

                       

                      http://vimeo.com/18007954

                      I had seen this video before, but again thanks of sharing it though. My main concern is, due to License issues i cannot use Nuke for all my Touch-up needs. Since Photoshop is cheaper, i have to set a pipeline to do the "Touch-up" (wire-removal) for all my 32 bit files in Photoshop.

                       

                      Any ideas are highly appreciated.

                      • 9. Re: "wire removal" on a 16Bit (Half float) EXR file format
                        prashanthvs Community Member

                        Hi Chris

                         

                        First: you seem completely confused between bit depth in files and documents.

                        In Photoshop: 8 and 16 bit/channel are integer formats, while 32 bit/channel is a floating point format.

                        EXR stores data as 16, 24, or 32 bit/channel floating point - which Photoshop needs to open into it's 32 bit/channel mode.

                        are you saying that Photoshop 8,16,32 is exactly equal to EXR's 16,24,32 ? (please correct me if am wrong)

                         

                         

                        Converting from 32 bit/channel floating point to 16 bit/channel integer lost any overranged values, and may have applied toning to the image (making the process non-reversible).  Your exposure changes added a lot of quantization, so you probably have 7 bits or less of actual image data left.

                        i totally agree on this concept. But can you please read once my reply which i just posted to "Mariam Driscoll" and let me hear your thoughts.

                         

                         

                        >> can you guys (professionals) confirm whether this process is safe.

                        I can confirm that what you describe is NOT safe.

                        Totally agree with you here. So, can you suggest me a perfect workflow were i need to "Remove Wire on a  32 Bit file using a Photoshop's "Healing tool" which works only on 16bit file"

                         

                        You really need to learn more about the formats and tools before trying something like that.

                        On the Nuke side, according to my understanding on 'formats' everything is reading/manipulating/writing perfectly. But i don't understand why under Photoshop, a 16Bit(half float) file is read as a 32Bit file?

                        • 10. Re: "wire removal" on a 16Bit (Half float) EXR file format
                          Chris Cox Adobe Employee

                          No, I'm telling you that it is NOT the same.

                           

                          EXR is usually floating point.

                          Photoshop has one floating point mode (32 bit/channel) and 2 integer modes (8 and 16 bit/channel).

                           

                          You *are* losing image data/quality when converting and using extreme exposure settings.

                           

                          The healing brush doesn't work in 32 bit/channel - that's it.  There is no workaround using that particular tool. You can use the clone stamp or other tools that do work in 32 bit.

                           

                          16 bit float reads into Photoshop's 32 bit float because that's the only thing it can do and preserve data.  Photoshop doesn't have a 16 bit float mode.

                           

                           

                          You really need to learn more about file formats, encodings, and applications before taking on something like this.

                          • 11. Re: "wire removal" on a 16Bit (Half float) EXR file format
                            prashanthvs Community Member

                            i will post a EXR file as an example shortly which should communicate better.

                             

                            thanks for your answers

                            • 12. Re: "wire removal" on a 16Bit (Half float) EXR file format
                              Chris Cox Adobe Employee

                              Posting the EXR file won't help.

                               

                              You need to understand the operations that you are using (scaling, toning, etc.) and how they damage the values in your image.

                               

                              You have a floating point image: if you want to retain the full quality in that image, do NOT convert it to 8 or 16 bit/channel.