Evangelos Achillopoulos
Well-known member
- Joined
- Apr 10, 2007
- Messages
- 558
- Reaction score
- 0
- Points
- 0
- Location
- Athens, Greece
- Website
- www.motionfx.gr
Many of you have herd about the differences in the two workflows while Color grading.
Because there is a quite big debate of which is better, I will try to show the differences in both workflows with a visual representation of the two.
The first figure (Fig 1.) is the Film target Log workflow that needs a calibrated display device and a custom Viewing 3D LUT specific for your display device to work during Color Grading. In Step 1 is applied the LUT and we see the RAW image not balanced. Notice the dark gray trapezoids that define the different headroom’s in color spaces of the display device and the RAW image files.
In Step 2 we have applied a Color Grade and image is viewed balanced. In this step a part of the image is outside of our display device viewing capabilities so we don’t see it.
In Step 3 we remove the Viewing 3D LUT and we render the image to a Log *.DPX format. So as we see, the resulted file has the full range just reallocated within the Cineon headroom.
In Step 4 we using the resulted Master *.DPX files and a set of Conversion 3D LUT’s, to convert the Film targeted Log image to the Linear color space of REC709 and DCP XYZ. During the conversion process, a part of the image information is removed in order the result to fit in the respective color space. In film deliverables we directly print on negative since the Master files are ready to be printed since we initially used a real Film target to Color Grade the images.
The second figure (Fig 2.) shows the REC709 workflow that also needs a calibrated REC709 display, not needed though a special 3D LUT. In Steps 1, 2 we do the some thing as with the Log workflow, we Color grade and balance our image to get exactly the same perceptual result as with Log, on our monitor.
In Step 3 we render our image in a Video format and we erase the part of image that is not visible. That is because of contrast... highlights and dark areas have information heavily compressed or even erased because they have to fit in 10bits instead of 12, 14bits or even 16bits of the original RAW images.
In Step 4 the image is ready for REC709 delivery, but for DCP we need a conversion LUT. The conversion LUT just stretches the image to fit, because our Master image is crafted for REC709 originally and doesn’t has that little extra to cover all the color space of DCP, so we sacrifice the maximum performance of DCP projection. A similar thing happens with the Film deliverables, where the problem is bigger though and as a result we have that too much video look in our image.
In figure 3 we have on the left side the Log results and on the right the linear:
As we can easily understand and to verify my facts, I challenge you to try to Re-Color Grade a final REC709 image and on the same time to do the same with a Film target Log final image. You will immediately realize that with REC709 you can't go far and with Film Log you can even restore the contrast totally like the image had never Color Graded.
That fact verifies what I'm claiming. And that fact gives Film Log workflow the upper hand in resulted quality in all types of deliverable's and makes your Master 4K images to be totally future proof because you don't throw almost anything.
The REC709 workflow has an inherit handicap, of less Gamut (color space) and less Headroom (bit depth), which Film Log workflow hasn't.
So Film Target Log workflow is way better than REC709.
Off coarse it needs custom LUT's... but nothing in life comes easy...
Disclaimer:
* For the sake of the argument and to make easier the build of graphics, I represented color spaces in the form of trapezoids, that isn't correct because I had to use cubes. Also the magenta is not the only thing missing in the different color spaces, but that was easy to be used for the representation.
Because there is a quite big debate of which is better, I will try to show the differences in both workflows with a visual representation of the two.
The first figure (Fig 1.) is the Film target Log workflow that needs a calibrated display device and a custom Viewing 3D LUT specific for your display device to work during Color Grading. In Step 1 is applied the LUT and we see the RAW image not balanced. Notice the dark gray trapezoids that define the different headroom’s in color spaces of the display device and the RAW image files.
In Step 2 we have applied a Color Grade and image is viewed balanced. In this step a part of the image is outside of our display device viewing capabilities so we don’t see it.
In Step 3 we remove the Viewing 3D LUT and we render the image to a Log *.DPX format. So as we see, the resulted file has the full range just reallocated within the Cineon headroom.
In Step 4 we using the resulted Master *.DPX files and a set of Conversion 3D LUT’s, to convert the Film targeted Log image to the Linear color space of REC709 and DCP XYZ. During the conversion process, a part of the image information is removed in order the result to fit in the respective color space. In film deliverables we directly print on negative since the Master files are ready to be printed since we initially used a real Film target to Color Grade the images.
The second figure (Fig 2.) shows the REC709 workflow that also needs a calibrated REC709 display, not needed though a special 3D LUT. In Steps 1, 2 we do the some thing as with the Log workflow, we Color grade and balance our image to get exactly the same perceptual result as with Log, on our monitor.
In Step 3 we render our image in a Video format and we erase the part of image that is not visible. That is because of contrast... highlights and dark areas have information heavily compressed or even erased because they have to fit in 10bits instead of 12, 14bits or even 16bits of the original RAW images.
In Step 4 the image is ready for REC709 delivery, but for DCP we need a conversion LUT. The conversion LUT just stretches the image to fit, because our Master image is crafted for REC709 originally and doesn’t has that little extra to cover all the color space of DCP, so we sacrifice the maximum performance of DCP projection. A similar thing happens with the Film deliverables, where the problem is bigger though and as a result we have that too much video look in our image.
In figure 3 we have on the left side the Log results and on the right the linear:
As we can easily understand and to verify my facts, I challenge you to try to Re-Color Grade a final REC709 image and on the same time to do the same with a Film target Log final image. You will immediately realize that with REC709 you can't go far and with Film Log you can even restore the contrast totally like the image had never Color Graded.
That fact verifies what I'm claiming. And that fact gives Film Log workflow the upper hand in resulted quality in all types of deliverable's and makes your Master 4K images to be totally future proof because you don't throw almost anything.
The REC709 workflow has an inherit handicap, of less Gamut (color space) and less Headroom (bit depth), which Film Log workflow hasn't.
So Film Target Log workflow is way better than REC709.
Off coarse it needs custom LUT's... but nothing in life comes easy...
Disclaimer:
* For the sake of the argument and to make easier the build of graphics, I represented color spaces in the form of trapezoids, that isn't correct because I had to use cubes. Also the magenta is not the only thing missing in the different color spaces, but that was easy to be used for the representation.
