I would like to share parts of one and my answers because i think it is a very interesting subject.
Here we go:
I just read your helpful review of the Sony F65 a while back which was very useful. In it you mention that the picture from the F65 can look somewhat videoish compared to, say, an Arri Alexa.
...
We have been amazed by the images but I have to say that I agree with you: 'straight out of the box' the images do look a bit more videoish than the Alexa and we have been trying to analyse why that is and the best way to reduce that videosh look in post production without reducing the quality of the image.
...
1. The image is very clean. But it pains me to add noise to the image (it goes against the grain of everything one has strived for in digital cinematography - if you forgive the pun).
2. The image is very sharp and detailed. This is an interesting point because I find the amount of high frequency detail quite filmic in a funny sort of way. It is the first digital camera that allows me to really see the fine texture I am used to from 35mm and 70mm film whereas the Alexa can feel like it is missing this texture compared to film and the F65. HOWEVER it does look to me like there is a tiny amount of sharpening being added in the demosaicing process although Sony tell us not. Anyway there is no way to reduce sharpening in the RAW conversion.
3. There is certainly no issue with the camera's colour response (as I have found with various iterations of the Red camera).
Ideally we would like pictures with the organic feel of the Alexa but with the high frequency detail of the F65. Any tips?
My answer follows:
First of all there is a miss conception that the grain reduces the sharpness or the detail when the reality is the opposite. A healthy dose of grain enhance the perception of sharpness for a simple reason, but to make it clear let me go back a little to the beginning of the story:
i heard many people arguing: since the eye acuity limit is on about 0.5 arc second a 2k footage projected in an average cinema reach the limit of what one can perceive.
Using this theory (greatly supported by the camera manufacturers because it suits their commercial strategy) 4k has more resolution than our eye and we will never perceive the difference between, say, 3k or 4k. The extra resolution will help to do postproduction and VFX.
But this is only true if you consider the eye as a camera, i mean, if you consider the eye as a static element. But the eye is continuously moving and "scanning" at very high speed, so the effective resolution of the eye is way much higher than the theoretical "2k at 10 meters on an average screen size" concept.
For a very simple further reading about the eye resolution see here: http://www.clarkvision.com/articles/eye-resolution.html
One of the big limitations of the digital systems compared to the film is that in the sensor all the photosites are always in the same place whether in the film the image data is made of grains that are randomly placed and randomly move from frame to frame greatly multiplying the "real" resolution.
Now it is easy to see that in a a limited media (lets say HD) how an added grain would enhance the perception of sharpness even if artificially. Ultimately the perceived sharpness lays in the domain of the perception what it is a very complex scenario.
Now one of the issues of applying noise to a footage is that many times the material looks bad or low quality.
This happens because of a limitation of how the grain is applied to the footage. Let me clarify this:
When i add grain or noise to a footage all the elements of the image will be noisy. But the real film doesn't work like that. In a real film the noise is present only in the mid tones and not in the blacks or whites. Simple: in a real film the black areas are non transparent, since no light is passing through there is no way to see the grains that creates that black. In the other hand the whites are transparent areas so there is no physical grain.
Normally to apply grain over the whole image not only makes the image look poor but more "videoish" since only the video has noise in the blacks (even more critical than the whites because the eye is much more sensitive to the shadows).
Happily it is very easy to do that in the digital world too.
In the first attachment called colour tree you can see as an example a tree created in the FX room of Apple Color that apply selectively grain in the area of your choice. More than that, you can create several branches of the tree and apply different characteristics of the grain in several areas given an even more natural feel and look.
Of course you can do this kind of treatment with a variety of software packages and in many different ways, this was only an example.
The other aspect of the filmic look is the fact that the cinema screen is dim, very dim (the SMPTE recommends a screen luminance of 16 foot lamberts for an open gate, no film loaded, what yield to about 14 foot lamberts with the film loaded or about 48 candelas per square meter). For this reason the cells responsible for the colour vision in our eyes (cones) are lightly stimulated while a big part of the vision comes from the rods. That means that the cinema looks desaturated, not because it is but because our eyes can not see the colours enough in a standard cinema.
One of the big differences between the Alexa and the F65 is, not only how saturated are the colours but the distribution of that colours over the spectra. Sony had always developed their cameras to have a very punchy look.
In the other hand it is absolutely true that as it is for today you have no control over the sharpness of the debayering. I hope this function will be added pretty soon by the diverse software manufacturers but now we can only add a soft blur or mist to the footage in postproduction or even better to use a flavour of a softening filter during shooting (a black pro mist could be a good choice).
One has to live with this tradeoff, there is no other solution :-(
on his reply:
I agree that Sony video cameras have traditionally had a very punchy look - but I assume this is largely down to the video processing circuits that handle colour matrix, gamma, detail etc. With the F65 in RAW mode we are dealing with a data camera with - one would think - relatively little processing going on except for some very mild compression and then debayer. I've never been convinced by the Sony colour matrixes but I assume these can be avoided when shooting RAW if using ACES coordinates or if grading in the native S-gamut and re-mapping the colours in to the colour space in which you are grading. Also, the gamma curve can be defined by the colourist who can apply more of a filmic 'S' shaped response curve. So we are left with debayer and noise.
So, I'm thinking that apart from the debayer, It ought to be possible to make the F65 look like a detailed Alexa by adding grain, bespoke gamma curve and being careful with colour space conversions.
my point about that:
Dont't be fooled by the word RAW. The capture and transformations previous of the digital stage (sensor, bayer pattern filters, amplifiers, AD converters and many other associated electronics) are 100% analogic and has their own characteristics varying from camera to camera. There are a LOT of transformations previous to the digital stage.
The debayering process in not native of the format, what i mean is that each software house creates their own algorithms for the demosaicing (pablo, davinci, sony, etc). In fact one of the beauties of the RAW format is that in the future one could get a much better quality over the old material. Today the quality of the demosaicing is limited by the computing power and the algorithms used. Basically it is a trade-off between a minimum desired quality and the time that it takes to process the material.
I can assure you that most of the debayering uses some kind of sharpening.
Of course this is not the end of the story.
If you like to write to me you can do it at martinchab@gmail.com
No comments:
Post a Comment