RAW files must be debayered before they can look like an image. With compressed RAW files, one must first decompress the file and then debayer it. The application that does this is called a raw processor. DaVinci Resolve is one example of an application with a built-in raw processor. It stands to reason that Blackmagic Design would have taken great pains to ensure Resolve worked well with their cameras. So, is Resolve the best raw processor for the Blackmagic Pocket Cinema Camera?
That’s what this article will try to answer. We’ll compare compressed CinemaDNG frames (recently made available by John Brawley here) in three applications:
- DaVinci Resolve Lite 10
- Adobe Camera RAW, via Adobe After Effects CC
- RAW Therapee 22.214.171.124
In a few days, Adobe will release native editing of CinemaDNG in Premiere Pro, though it remains to be seen if compressed DNG will be supported as well. It should, though, because Blackmagic Design haven’t reinvented the wheel. The lossless compressed files open in all of the above applications. Therefore, we can safely assume that they have remained within the specifications of CinemaDNG, and have not included any proprietary compression.
This article builds on the following, so read it first if you haven’t already:
What makes a good raw processor?
A camera sensor has pixels. If it’s a bayer sensor (as most cameras are, including the Blackmagic Pocket Cinema Camera), then it also has a Color Filter Array (CFA) that tells it which pixel is red, blue or green. In addition to the CFA, a sensor might also have an optical low-pass (OLP) filter, and other filters. The Blackmagic Cinema Cameras don’t have OLP filters, so they tend to display greater moire and aliasing.
Now, the distance between each pixel (not as simple as it sounds) is different for different sensors. E.g., the sensor in the Blackmagic Cinema Camera is totally different from the Pocket Camera. This is why most camera manufacturers also make raw processors:
- Canon Utilities RAW image converter
- Nikon Capture NX 2
- Blackmagic Design DaVinci Resolve
- Sony RAW Viewer
- Arri Arriraw Converter
- Red Redcine-X Pro
Nobody knows their sensors better than they do. Most third-party raw processor developers must first have access to the camera sensor (so they must typically wait till the camera is in the market) and then study and test it thoroughly. From this, they derive the appropriate algorithms for each camera. The point is, debayering isn’t a simple process, and is as much art as it is science. For this reason, you’ll never find two raw processors agreeing on any image.
So, a good raw processor must:
- Have an algorithm custom-made for the exact camera sensor.
- Have good workflow options (otherwise you’ll have headaches like the Magic Lantern RAW to DNG problem).
- Must offer important image processing tools for raw, like sharpening, LUTs, custom settings, camera metadata support, etc.
- Account for many scenarios – flare, aberrations, distortions, filters, noise, varying light, and even motion (for video). This means, a raw processor must also ‘know’ about the likely lenses the sensor is going to be paired with.
The more information the programmers have, the more accurate it can be.
Who should test raw processors?
How involved should you get with raw processors? It depends. How fanatical are you about image quality? Ignorance is bliss, so you could opt for ‘whatever’, and if that makes you happy then that’s all that matters.
On the other hand, these individuals will find it beneficial to test the available options:
- DPs or Camera Assistants (how many productions can afford to tag along a DIT?)
Ultimately, ignoring raw processing is ignoring a large part of what makes raw, raw.
What are we testing for?
What do we hope to achieve? I can tell you straight away that none of the processors we are testing are bad in any absolute sense. That’s the problem with raw processing, the variations are either so minute or so off-on-a-tangent it becomes a matter of taste. Subjective opinion is a large part of the process.
From a workflow perspective, these are the factors I’m going to consider:
- Default state: How does the processor manage camera settings?
- Film mode: How does the processor manage the ‘film’ mode?
- Debayering algorithms: How many options do you have?
- Noise performance (Very important)
- Sharpening (Very important but also frustratingly subjective)
- Overall impression of color (highly subjective)
- Workflow and output codecs
John has made available five frames. I’ll be testing only three – frames 19, 40 and 60. 19 looks blown out (but is not, in film mode). 40 has skin, flare and motion blur. 60 has a lot of fine detail and areas of shadow that we can check for noise. Here are the three frames (compressed and resized JPEGs):
To minimize errors, I’ll export TIFFs from each application and then use Photoshop CS6 to compress to JPEGs. I’ll only provide crops if relevant. No changes in color will be made.
Stuff I’m not testing
- Color temperature – I won’t be varying the color temperature from ‘As shot’ (camera metadata). All these shots were at about 5550 K.
- Color space – I’m going to be choosing an output space of sRGB because that’s how everyone’s going to read this article. Because it’s raw, there is no baked in color space.
- Gamma – I’m going to be using Film gamma for the flattest image. The ‘stretchability’ of the image in grading is not dependent on raw processing, but on factors like dynamic range, color space, pixel electronics, sampling, etc.
- Default tint and sharpness – the Blackmagic Pocket Cinema Camera applies a default tint of 13.97 (rounded off to 14 sometimes) and a sharpness of 10 (algorithm not available) to every image. This is similar to the performance of the original BMCC.
Resolve 10 Lite and RAW Therapee are free. Adobe CC is not. All three applications are available for both Windows and Macs.
How fast can we get from A to B? When a DP shoots on set with a look based on Rec. 709, you’ll want to start from that. This is similar to what you’d get if you shot Prores.
RAW Therapee does three things differently:
- It increases the size of the frame from 1920 x 1080 to 1944 x 1104, and this changes the aspect ratio.
- It applies Auto levels by default. This causes the exposure to be set so the histogram sits perfectly between 0 and 255. Why? Because –
- It cannot read camera metadata at all, and you can see clearly that it cannot reproduce exposure automatically. Image 19 is somewhat exposed to the right and RAW Therapee knows it – but it doesn’t know the color space or the exposure.
After Effects has the following idiosyncrasies:
- It adds sharpness and noise reduction by default, and I had to turn it off. However, the Blackmagic Pocket Cinema Camera does add sharpening by default, so the images appear slightly sharper in Resolve.
- It also manages to add a tint of 14, and can read the white balance and exposure settings off camera metadata.
- It misses exposure but not by much, however the noise in the shadows become more visible. Here’s a comparison of the two histograms:
The red circle is a really big difference. ACR does not contain as much information in the shadows. To test this, I used the levels curve to stress test the images (just a fancy way of saying I took the values to the extremes), and the images from Resolve held out better in the shadows. Here’s the histogram comparison for each channel:
You can clearly see that Resolve manages to get more information in the shadows (black regions). ACR holds more green information in the highlights, though. So, when you want to get more information in the highlights, ACR is better. From this, I can only surmise that they are both equal, really. If ACR had a calibrated exposure tool designed for the Blackmagic Pocket Cinema Camera, the differences would be negligible.
In any case, judging a raw camera or raw footage by looking at its performance in Rec. 709 is hardly fair. The real comparison happens in the film mode. Let’s tackle this in Part Two.