Addl corrections

[jeremyselan]

page 27: "is loaded into the DI" -> "is loaded into the DI system" page 27: "to create the final the color appearance" -> "to create the final color appearance" page 27: "it is likely that in the overexposed image that large portions of the image was clipped to a constant maximum value" -> "it is likely that large portions of the overexposed image were clipped to a constant maximum value" page 27: "an S-shaped tone curve is utilized; crafted to precisely emulate" -> "an S-shaped tone curve is utilized, crafted to precisely emulate" page 27: can you give a horizon where feature film DI will no longer be concerned with film print distribution? page 27: last paragraph, not super clear on relative merits of DPX vs EXR for CG elements, what about half float in DPX? Also, EXR is often not preferred for DI as some of the compression methods make real-time playback more difficult?

page 28: "this often is send" -> "this often is sent" page 28: "and then a secondary process": that's a bit confusing, in grading "secondary color correction" is typically used as a synonym for "selective color correction", i.e. changing the color of only a part of the image through a matte generated either from a pre-rendered mask (typical in CG animated movies), a hand-drawn geometry shape ("Power Window" in DaVinci speak) or a keyed matte

page 29: "for digital projectional" -> "for digital projection" page 29: "The director, producers, and other creatives will be present at this process" -> I would explicitly mention the DP as a (desired) participant in the DI process, since he/she has typically the most to say about the desired look of a movie page 29: "as the colorist already familiar with the final look of the material" -> "as the colorist is already familiar with the final look of the material" page 29: a couple of notes about stereo DI: on top of color, lots of stereo-specific "stuff" happens there, in particular image alignment and "dialing in" of stereo separation (may not be relevant here?). Also, maybe add a note that the dimness of stereo 3D is an undesired artifact of current display systems that will hopefully be remedied in the future? Any thoughts about DI for HFR/48FPS projection? Did anything special have to be done for The Hobbit?

page 31: "Physically plausible light transport renderer mechanisms such as global illumination yield natural results when given scenes with high dynamic ranges. Physically-based specular models, combined with area lights, produce physically plausible results with high-dynamic range data." The two sentences are almost identical. How about something like: "Physically plausible light transport render mechanisms such as global illumination yield natural results when given scenes with high dynamic range data. Specular lighting models, combined with area lights, also work best with HDR data." or something like that. Also, inconsistent use of "high dynamic range" vs "high-dynamic range" page 31: "with results that track identically to if the original render had been tweaked." -> "with results that match tweaks to the original render, a much more computationally expensive approach". page 31: "is most common in VFX and animation" -> "is most commonly used in VFX and animation to store floating point data." page 31: "Even still, it is common" -> "Even with the use of such techniques, it is common…" page 31: "lower-noise" -> "lower noise" page 31: "in computer-graphics" -> "in computer graphics" page 31: "it’s very difficult to avoid clipping" -> "it is very difficult to avoid clipping" page 31: "is that it’s often required" -> "is that it is often required" page 32: " as it’s much friendlier" -> " as it is much friendlier" page 31-32: Not sure I understand the sentences "On the downside, one consequence… to a RGB value of "1048576.0"" I don't really see the need for having UI elements in f-stops as a downside, and in fact is probably more intuitive to anyone who has a photography background? Maybe just remove the "On the downside" part?

page 32: "can be rolled-off" -> "can be rolled off" page 32: "of the scene-lighting" -> "of the scene lighting"

page 33: "of the motion-picture" -> "of the motion picture"

page 34: "such as motion-blur" -> "such as motion blur" page 34: "in display-linear" -> "in display-linear space"

page 35: "There are some challenges with working with" -> "There are some challenges when working with" page 35: "While interpolatory filters" -> "While interpolating filters" page 35: "the results are visually pleasing and is suitable" -> "the results are visually pleasing and are suitable" page 35: this discussion seems to duplicate the first paragraph of

page 32, which discussed filtering in the context of rendering, although it is more detailed. Any way to combine the two, or have the filtering discussion under compositing refer to the section on rendering?

page 36: would it make sense to give a brief definition of the screen operator? That paragraph doesn't seem very helpful without a definition of the underlying operators. page 36: "Another issue to beware of" -> "Another issue to be aware of" page 36: would it make sense to mention that "NaN" and "Inf" are "out of range" floating point representations, and that working in "half float" makes those more likely to occur than when working in 32 bit of 64 bit float due to limited prevision available? page 36: when compositing in float, it is typical to use a "stop up / stop down" display LUT to verify that what you are doing isn't introducing artifacts in areas that aren't easily viewed with the current display transformation. Is this still valid when working in scene-referred, and what kind of display LUT would you use to reveal such artifacts? In fact, isn't this even more important than when working in log, since what you see on your monitor is an even smaller proportion of the original data than when working in log or displayed referred? There is some discussion of this in the context of matte painting on page 40, but it might make sense to also discuss it in the context of compositing. OK, this is discussed in detail on page 41 under "critical inspection", maybe have a forward reference to that section?

page 37: "there are just some cases even in modern compositing workflows where log-like spaces are still useful" -> "even in modern compositing workflows there are cases when log-like spaces are still useful" page 37: "Contrast is typically describes" -> "Contrast is typically described" page 37: "This is one of the reasons that DI color work is often preferred to work in log-like spaces" -> "This is one of the reasons why DI color work is often done preferentially in log-like spaces" page 37: "leads to nice gains in efficiency" -> "leads to useful gains in efficiency" page 37: "as these type of corrections" -> "as these types of corrections" page 37: "presuming the linearization was handled done to begin with" -> "presuming the linearization was handled to begin with"? not too sure what this is supposed to mean page 37: "between timing number necessary for plate neutralizations" Not sure that you have introduced the term "timing number(s)" before, how about "between color timing RGB offsets necessary for plate neutralization" or something like that? page 37: "I.e." -> "i.e" page 37: "Beware that if you bake in…high or low end" This sentence seems confusing to me, not sure how ti reword it?

page 38: nitpicking here, but if RGB represents light emitted by the pixel, then stating that "RGB channels can be negative" is a bit counter intuitive, since there is no such thing as negative light page 38: "it's prudent to only do" -> "it is prudent to only do"

page 39: "the texture painting is not be truly WYSIWYG" -> "is not truly WYSIWYG" or "cannot be truly WYSIWYG"

page 40: "it's possible to create" -> "it is possible to create" page 40: "such as dcraw" -> add reference to dcraw website? page 40: is discussion of trickiness of filtering scene referred data (from rendering and compositing sections) applicable to mipmap generation? What about reconstruction filters for texture sampling used by shaders?

page 41: "during DI is possible that" -> "during DI it is possible that" page 41: "in unanticipated direction" -> "in unanticipated directions" page 41: [21] Can this be rephrased as a bit less of a "dig" at DI? Also, I would stick to actual references as bottom notes? page 41: "for it's all too easy" -> "for it is all too easy" page 41: "when viewed side to side" -> "when viewed in the context of the entire sequence"? page 41: "flat black portions of the image should be avoided - lest they reappear in DI" -> not exactly sure what this means? How about something like "crushed shadows, where an entire section of the image clamps to the minimum black value, should be avoided as they will often "pop out" when any color grading is applied in DI page 41: "and often help forensically track down" -> "and often help to forensically track down" page 41: "as a final quality-check" -> "as a final quality check" Note sure the sentence about HSV histograms is very clear: how does one use an HSV histogram to quality check images? Why are those better than histograms computed in a different color space? page 41: "interpolatory filters" -> "interpolating filters" page 41: "moire" -> "moiré" page 41: (examples: cloth/wave patterns) -> (such as cloth or wave patterns) page 41: "overly-sharp" -> "overly sharp" page 41: "it's critical to quality check" -> "it is critical to quality check"

page 42: "aces files are stored" -> "ACES files are stored"? page 42: "defines scene-referred working space" -> "define a scene-referred working space" page 42: "the viewing transform necessary" -> "the viewing transforms necessary" since there are both the RRT and ODT? page 42: "a series of input device transforms (IDTS)" -> "a series of Input Device Transforms (IDTS)" page 42: "manipulating the aces data" -> "manipulating the ACES data" page 42: "(CTL); contributed by Industrial Light & Magic" -> "(CTL), contributed by Industrial Light & Magic" page 42: "and is thus suitable for baking into 1D/3D LUTs" -> "and is thus suitable for baking in 1D/3D LUTs" (otherwise it sounds like you are "baking CTL into 1D/3D LUTs"), or am I misunderstanding? page 42: "Unfortunately though," -> "Unfortunately," page 42: What are the primaries for the ACES working space? And how different / similar is the ACES working space to the scene referred space" typically used by large facility CG pipelines?

page 43: is there a way to have a higher resolution screenshot embedded in the PDF? it's a bit hard to see the difference (or lack of!) between the two apps showing the Kodak reference image page 43: "applied without the user awareness" -> "applied without the user being aware of it". Not only applications are guilty of this, media frameworks and operating systems as well (QuickTime / OS X!) page 43: building-blocks -> building blocks page 43: “and per shot looks” -> how are these specified?

page 45: “are an technique” -> “are a technique” page 45: the discussion on the performance advantages of LUTs may not be as relevant in modern systems, where it is often faster to evaluate a function rather than doing the memory / cache lookups for pre-computed values. I would suggest that a better justification for LUTs is to capture empirically measured data that cannot easily be expressed by a mathematical function, as well as to avoid the ambiguity of trying to express and evaluate a mathematical function in a platform independent way. page 45: “amenable into baking into” -> “amenable to baking into” page 45: “alternative formations are required” -> “alternative formulations are required” ? Not sure this is what is meant page 45: “Consider an analytical...significant speedup”: same comment as above, on modern architectures this may not be in fact a speedup, unless you are computing something very complicated (and how complicated can a function of a single color value be?). page 45: “parameterized” -> parametrized (parameterized or parameterised is UK spelling, which I have nothing against but is not what is used in the rest of the document) page 45: “can be accelerated”: same comment as before, main focus of LUTs these days may not be acceleration

page 46: “parameterized” -> “parametrized” page 45, 46: the quality of the figures embedded in those pages is pretty low (JPEG?)

page 47: “appropriate acceleration technique” -> “appropriate implementation technique” ? In 3D LUTs, it seems that capturing measured data is even more the focus than trying to pre-cache a computation: the memory footprint of a 3D LUT and the need for sophisticated interpolation means that you are even less likely to gain acceleration for something that could be described analytically (unless that function of 3 variables was incredibly complex and required lots of if/then statements as it varies across ranges of inputs, modern CPUs hate if/then statements) page 47: “when applied to alternative imagery” -> “when applied to different imagery” (the problematic imagery is not an option to the original imagery, it is just different) page 47: “through the color processing” -> “through the color processing pipeline” ? page 47: a small point, but maybe worth mentioning: 3D LUTs are are usually defined with “2^n + 1” data points per axis to help with interpolation calculations page 47: quality of figure embedded in PDF is low

page 48: “The world of color correction, particularly as it is handled onset, has a huge amount of variation.” -> There is huge variability in the way color correction is handled, especially when it is done at image capture time on a movie set. page 48. “scales and offsets. (The opposite is possible, of course).” -> “scales and offsets, since the desaturation operation must come last.” page 48: I’m not sure I “buy” the argument that the fact that CDL is under specified is its “biggest strength”: it seems to me that the fact that it is under specified is allowing people to abuse / override its intended meaning, but then it just becomes a pretty way to write numbers as a bit of XML? Could maybe be rephrased?

page 49: per-light color AOVs, etc. -> missing closing bracket. Also, may want to write out explicitly what AOV means (I’m not sure I know) page 49: “The minimum value which can be represented, ” -> “The minimum positive value which can be represented,” page 49: “if you’ve got a” -> “if you have a” page 49: “who created it (or detailed forensic analysis)” -> “who created it, or detailed forensic analysis”. -> Would be great to have some “forensic analysis” tools in opencolorio! page 49: “But beware these flags...” -> “But beware of these flags” Also, destruction of header metadata is a common problem to all image file formats, not just DPX: few applications are diligent in preserving header metadata, especially when the application is actually modifying the images instead of just copying them, it is often unclear how the header should be rewritten. Most image formats have fields that are defined rather ambiguously, and the data that is written in there is more typically described as “accepted practice” by the various applications that implement writing of those file formats.

page 50: “with a reference white point of 48cd/m2” -> isn’t that better said as “with a reference white luminance level of 48 cd/m2”? Isn’t “white point” usually used for the x,y coordinates that define the “color white”? page 50: “The intent of the X’Y’Z’ coding space is display-referred, such that the full color appearance of the theatrical imagery (such as any film emulation 3D-LUTs) are fully baked into the X’Y’Z’ image.” -> This sentence is a bit confusing to me. Maybe something like “The intent of the X’Y’Z’ coding space is to be display referred, so the full color appearance of the theatrical imagery (including any film emulation 3D LUTs) must be fully baked into the X’Y’Z’ image”. page 50: the link to the DCI Spec is broken, you get to the DCI site but it says the file is not found

page 51: “, it is often preferable to convert from spectral representations to RGB using camera inspired sensitivities, rather than an XYZ/matrix approach.” I’m not sure this sentence gives enough info to understand what is meant? page 51: “it’s important to note” -> “it is important to note” page 51: it is pretty hard to see the difference between the blackbody and daylight curves on the figure, should the figure be larger, or use a different line style for the two curves?