<div dir="auto"><div><br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">вт, 7 нояб. 2023 г., 21:46 Andrew Randrianasulu <<a href="mailto:randrianasulu@gmail.com" rel="noreferrer noreferrer" target="_blank">randrianasulu@gmail.com</a>>:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="auto">oiiotool --stats -v /dev/shm/1.exr Reading /dev/shm/1.exr /dev/shm/1.exr : 4096 x 2048, 3 channel, float openexr <div dir="auto">channel list: R, G, B chromaticities: 0.639999, 0.33001, 0.300004, 0.600003, 0.150002, 0.0599972, 0.3127, 0.329 compression: "zip" PixelAspectRatio: 1 screenWindowCenter: 0, 0 screenWindowWidth: 1 oiio:ColorSpace: "Linear" oiio:subimages: 1 Stats Min: 0.005371 0.006256 0.006348 (float) Stats Max: 117.000000 106.500000 104.500000 (float) </div><div dir="auto">Stats Avg: 0.746420 0.699325 0.674529 (float) Stats StdDev: 3.869811 3.841790 3.957525 (float) Stats NanCount: 0 0 0 Stats InfCount: 0 0 0 Stats FiniteCount: 8388608 8388608 8388608 Constant: No Monochrome: No</div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto">this is for gimp exported EXR</div><div dir="auto"><br></div><div dir="auto">cingg exported tiff says:</div><div dir="auto"><br></div><div dir="auto">Reading /dev/shm/1.tif /dev/shm/1.tif : 4096 x 2048, 4 channel, float tiff </div><div dir="auto">channel list: R, G, B, A compression: "zip" Orientation: 1 (normal) planarconfig: "contig" oiio:BitsPerSample: 32 tiff:Compression: 32946 tiff:PhotometricInterpretation: 2 </div><div dir="auto"> tiff:PlanarConfiguration: 1 </div><div dir="auto">tiff:RowsPerStrip: 1 </div><div dir="auto">tiff:UnassociatedAlpha: 1 </div><div dir="auto">Stats Min: 0.265371 0.266256 0.266348 1.000000 (float) </div><div dir="auto"> Stats Max: 1.000000 1.000000 1.000000 1.000000 (float) </div><div dir="auto"> Stats Avg: 0.630706 0.600794 0.582030 1.000000 (float) </div><div dir="auto">Stats StdDev: 0.221808 0.214916 0.202718 0.000000 (float) </div><div dir="auto">Stats NanCount: 0 0 0 0 Stats InfCount: 0 0 0 0 Stats FiniteCount: 8388608 8388608 8388608 8388608 </div><div dir="auto">Constant: No Monochrome: No</div><div dir="auto"><br></div><div dir="auto">====</div><div dir="auto"><br></div><div dir="auto">example usage from </div><div dir="auto"><br></div><div dir="auto"><a href="https://openimageio.readthedocs.io/en/latest/oiiotool.html" rel="noreferrer noreferrer noreferrer" target="_blank">https://openimageio.readthedocs.io/en/latest/oiiotool.html</a><br></div><div dir="auto"><br></div><div dir="auto">wow, so oiiotool in itself a bit like imagemagick :)</div><div dir="auto"><br></div><div dir="auto">might make preparing some image sequences easier ...</div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto">also, guess we better to note in manual that even if cingg outputs fp32 exr/tiff values there are normalized to 0-1.0f as far as we can see.</div></div></blockquote></div></div><div dir="auto"><br></div><div dir="auto"><a href="https://docs.krita.org/en/general_concepts/colors/bit_depth.html" rel="noreferrer noreferrer" target="_blank">https://docs.krita.org/en/general_concepts/colors/bit_depth.html</a><br></div><div dir="auto"><br></div><div dir="auto">Also, Krita docs says about 16 bit per ch integer vs 32bit per channel fp without values over 1.0f</div><div dir="auto"><br></div><div dir="auto">====</div><div dir="auto"><br></div><div dir="auto"><dt>16bit per channel</dt><dd><p>One step up from 8bit, 16bit per
channel allows for colors that can’t be displayed by the screen.
However, due to this, you are more likely to have smoother gradients.
Sometimes known as “Deep Color”. This color depth type doesn’t have
negative values possible, so it is 16bit precision, meaning that you
have 65536 values per channel.</p></dd></div><div dir="auto">=====</div><div dir="auto"><br></div><div dir="auto"><dt>32bit float</dt><dd><p>Similar to 16bit float but with even higher
precision. The native color depth of OpenColor IO, and thus faster than
16bit float in HDR images, if not heavier. Because of the nature of
floating point type variables, 32bit float is roughly equal to 23-24
bits of precision per channel (16777216 values per channel in the 0-1
range), but with a much wider range (it can go far above 1), necessary
for HDR/Scene-referred values. It is also known as ‘single floating
point’.</p></dd><dd><p>====</p></dd><dd><p><br></p></dd><dd><p>so I guess this mean even WITHOUT going above 1.0 fp output format still more accurate than 16bit per channel integer tiff or png. But I am bad at math for PROVING it. :(</p></dd><dd><p><br></p></dd><dd><p><br></p></dd><dd><p><br></p></dd><dd><p><br></p></dd></div><div dir="auto"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
</blockquote></div></div></div>