Color Science: The photochemical look on digital captured format.

PLESAE NOTE! This webpage is not finished. It will be updated as soon as my thesis on color science is finished.

The color conversion in a nutshell

The complex Kodak Vision3 5219 500T that I have developed, expects ARRI LogC Wide Gamut. The 5219 that I used side by side with a digital cinema camera, was scanned at the highest possible quality with ARRISCAN in DPX 16Bit Cineon. In my testings I have also been using a BMPCC4K, and it worked beautifully when using Emotive Color to accurately match the ARRI Alexa's Log-C color response and tonality (below is a side by side). The LogC footage is then converted the the Kodak Vision3 5219 color science in Cineon log, and then printed to a custom Kodak 2393 FPE in a Rec. 709 (BT. 709) space. In my testings, both the ARRI Alexa and BMPCC4K performed equally amazing and accurate with disposable differences. Due to the flexibility of the transform, it can also be used in Dolby Vision and HDR10 workflows for example with a trim pass or natively. 

Below we are some matched shots from digital captured format matched with analogue motion picture film


I used the two sample shots from JSFILMZ below to prove how equal both camera's perfom.

ARRI Alexa and BMPCC4K (with ARRI Alexa color science). Both converted with the same technical transform to match 5219.


In this comparison we can observe the beautiful the color separation when the transform is applied to the BMD Film gen4 (with Emotive Color).


Film Grain

Grain emulation is quite difficult to replicate. But after a lot of research and fiddling around, I developed a complex and precise film grain emulation that responds to color density, and is generated in real time. It was developed based on a scientific analysis of the the Kodak Vision3 5219 500T (35mm) grain structure. It deeply embeds the grain into each color channel, while also responding accordingly to the dyes density. The same principles apply to other negative film stocks, but with different intensities and responses. 


Converting the emulation to 16mm film would either have be done via a 2K recording (it is now calibrated to 4K input), so we get a native conversion of the grain size. Or the entire emulation (also including film halation) needs to be scaled by a relative factor of about 2x. We could of course also lower or raise the scaling factor depending on the size of the gate.


Here we can see the color channel separated images, and a final combined image where all three color channels are mixed. This demonstrates that all individual RGB channels have a different intensities and grain structure.


Both left and right screengrabs are from a 2K lossless source. The left screenshot is from Wonder Woman 1984, and on the right we see the complex film grain emulation (v2.0) on digital captured format. To point out accuracy, we can see that the grain also organically generates the blue and yellowish noise patches in some conditions.