Resolution, sharpness and colordepth.

[h-munster]

At the request of other posters, this thread is forked from the Being Cinematic thread. There was an "off-topic" discussion developing regarding the look of an image when the viewing system has less resolution than that of the camera.

One poster seemed to assert that the viewing sharpness will be identical from all camera resolutions that are greater than that of the viewing system (assuming all other variables are the same).

I linked articles by two experts, one from Panavision's John Galt and the other (PDF) from Arri's Dr. Hans Kiening. Galt downplays the importance of shooting hyper-HD resolutions in regards to viewing sharpness, while Kiening makes a case for keeping resolution as high as possible in all image processing steps, from shooting to transferring (film) to viewing.

I also made a point that is negelcted in almost every resolution discussion -- resolution has a strong affect on color depth. Basically, the higher the resolution, the greater the color depth. This fundamental photographic principle applies to all photographic medium, and it can affect the final look on a lower resolution display, if the bit depth is high enough.

For the most part, color depth is a product of resolution and bit depth. Suprisingly, it only takes a small increase in resolution to make the color depth rise astronomically.

More info on how resolution affects color depth can be found here.

 

[Jordan Scott Price]

Is his name really John Galt? :shocked: I've never heard of him before in relation to film.

 

[h-munster]

Is his name really John Galt? I've never heard of him before in relation to film.

He has a somewhat illustrious career in film. John Galt is a cinematographer, and he is also Panavision's Senior Vice President of Advanced Digital Imaging. Among other things, he was responsible for the creation of the F900 and for the development of the Genesis camera.

 

[Ryan Patrick O'Hara]

Yes, you are right about most things, but I don't think it really makes a difference when comparing such high formats to the ultra low distribution of DVD.

More information is simply better to start. It helps keep things nice when doing optical effects/printing, multiple generational dups, etc.

Dynamic range is also greater by a perceived effect of higher resolution. Meaning 52__ will appear to have increased dynamic range over 72__ kodak stock.

My point wasn't that there are not any advantages to shooting 15 perf IMAX 65mm or 5 perf 65mm.... sure there are... for all the things you listed and more. It's best to originate on a larger medium... BUT I will say that I'm not sure there is an advantage of shooting 5-perf 65mm over 35mm when viewed on a 300x704 standard def DVD or 480x704 standard def DVD.

At that compression and tiny resolution, I don't know if it would be worth it.

The only thing i can think of is how spherical lenses on 65mm could benefit over anamorphic 35mm at open apertures. I've yet to be sold on clarity and sharpness of anamorphics wide open.



So I agree that originating on larger formats, yields greater information, which yields greater control and variations. That's good.

 

[David G. Smith]

....makes a case for keeping resolution as high as possible in all image processing steps, from shooting to transferring (film) to viewing.

I also made a point that is negelcted in almost every resolution discussion -- resolution has a strong affect on color depth. Basically, the higher the resolution, the greater the color depth. This fundamental photographic principle applies to all photographic medium, and it can affect the final look on a lower resolution display, if the bit depth is high enough.

But isn't that the caveat?

Doesn't modern workflows really ameliorate a good deal of the advantage that uber high resolution origination COULD provide. Let's be simple about it and use the difference between shooting 35mm film and 65mm film. There is no doubt that if you compare 65mm film to 35mm film, both finished photo-chemically to there respective formats, that the 65mm film will be superior, all other variables being equal. Now you go down the food chain and do 65mm and 35mm, finished photo-chemically, but with the 65mm film being presented as a 35mm reduction. Even if the reduction process is done from the camera original negative, which I doubt anyone would do today, are you saying that the 65mm originated film will look better? Even if you did everything else the same, interpositive to internegative to release prints on the same stock and equipment, I don't think that the 65mm originated material would be perceivably superior.

Lets go further down the food chain. 65mm film and 35mm film to 4K scan, to 4k finish, at what ever is the best digital format possible, to 4k presentation. Ok, I could see the 65mm film holding up better, but wouldn't the color depth be set by the capabilities of the scan? Since they are both now being scanned at the same resolution, 4K in this example, where is the advantage?

And further down the food chain, doesn't thing get worse and worse, and the advantages of the uber high resolution origination kind of peak at a point, especially in regards to color depth?

In the real world, using real work flows, the advantages of higher resolution origination is an advantage only if the entire work flow supports that advantage. To me, that was one of the main findings of the Zacuto Shootout. The colorist made quit a point about the 35mm film being much easier to grade, and having much better color depth, but that it was not because of the HDDSLRs resolution, but from the HDDSLRs sh*tty origination codec.

I believe that once an image becomes ones and zeros, there are just so many other factors that need to be taken into consideration than JUST resolution.

 

[Ryan Patrick O'Hara]

I guess to sum up my belief regarding the discussion:

When going to 480x704 or less... whether one is shooting on 35mm, 65mm or 100,000mm really isn't a big deal. All those advantages of a large format origination we discussed should be in play at the 35mm stage if going to 480x704.

Now if the film was finalizing on 35mm, than if money was no object, shooting 65mm, doing all the color timing, optical effects, and etc and then printing to 35mm would yield better results.


I guess it's just a game of origination versus deliverable. That was how this whole discussion started... it was never about bigger formats being worse or not good... it was about DVD viewing. At some point, originating large is good, but originating larger becomes pointless.

 

[Lammy]

it was never about bigger formats being worse or not good... it was about DVD viewing. At some point, originating large is good, but originating larger becomes pointless. .

Yeah I don't think we've reached that point yet. It's pretty subjective on whats considered "pointless". But of course, we can't just compare resolution or sharpness, but the whole different set of diferent tool kits that bring out this look. And as artists carving an image, its only as pointless as the people utilising it think so.

I might be able to tell IMAX compared to 35mm on DVD, but others might not.

Most people I've spoken to think 28 Days Later on DVD looks like 35mm film. I don't think so.

Red MX certainly has bigger pixels and colour depth... can't say I'm a huge fan of the highlight clipping or colorimetry though.

120 Film or really detailed pencil art?

 

[j1clark@ucsd.edu]

Lets go further down the food chain. 65mm film and 35mm film to 4K scan, to 4k finish, at what ever is the best digital format possible, to 4k presentation. Ok, I could see the 65mm film holding up better, but wouldn't the color depth be set by the capabilities of the scan? Since they are both now being scanned at the same resolution, 4K in this example, where is the advantage?

The higher resolution 'enhances' the representation of the image area. It does not change 'dynamic range' in the least.

Say you have an area that has a 'slowly' changing color/value. Sample it at 10 points, you have '10 samples of the variation'. If your digitizer can represent the 'changes' between those 10 points, then you will have captured the 'dynamic range' of the area.

If on the other hand the 'dynamic range' of your digitizer maps the intensity levels to the same 'value', because of limited dynamic range... you will have 10 pixels which have the same 'color/value'...

For example, if a gradient is made, and your digitizer only has 256 values, and the change from pixel to pixel is less than 1/256 of the range of the digitizer, you wil have banding whether you digitize to 1024 or 4096 pixels along a horizontal line...

If on the other hand, one has high dynamic range, and high spatial resolution, one can get better representation by the higher sampling rate.

The whole issue of 'viewing' with regard to 'resolution' and 'dynamic range', is intimately tied to 'viewing distance' as well...

images shot with limited resolution, and limited dynamic range, end up falling apart when viewed 'blown up'... no matter the medium... it just so happens that 'silver film grain', aka 'noise' is more aesthetically pleasing than current 'digital noise' in the form of excessively 'enlarged' digital images...

I've not looked into this in quite a long time, but I recall the human vision system was 'good' for detecting like 16 levels of gray in a 'large field' view, and perhaps up to 64 levels when the focus was narrowed to a limited field...

Part of this sort of 'reasoning' was put in to the various compression techniques used when humans would be the viewers of the media...

The problem of course is when people want to use a 'compressed with human vision considerations' media, in a 'process' where it is more desirable to have media that is not so compressed, the result is the dynamic range has been compressed, and as such may be more limited than a medium, such as 'film', which does have it's own compression, but is not 'designed' for human perception.
(That is 'film' responds non-linearly to 'light', and 'compresses' at both ends of the scale, low light, as well as high light... and in oder films, use to go into 'inversion' mode if given too much light... aka. Solarization...)

When I was more 'in' with things at 'fuji' one of the methods that their sensors used to gain better highlight response was to have two sensor cells where one was less sensitive, and so, could respond without saturation to higher levels of 'light'. So 'Cell 1' would repsond to low/mid light, "Cell 2" would respond to 'mid/high', and the result was 'increased dynamic range'... but I digress...

To sum up, for me 'increased resolution' only gives better 'dynamic range' representation, when the underlying sensor/digitizer has high dynamic range.

One can of course fake to some degree the human vision system into believing there is 'more resolution', or more 'dynamic range'... but ultimately it boils down to 'more for more'...

 

[j1clark@ucsd.edu]

I guess it's just a game of origination versus deliverable. That was how this whole discussion started... it was never about bigger formats being worse or not good... it was about DVD viewing. At some point, originating large is good, but originating larger becomes pointless.

Since the digital intermediary process has pretty much become 'standard', and the process has typically been 2K, that is not so far of HD's 1920x1080, that one can really talk about 'spatial' differences. The biggest difference is in the dynamic range of the digital sensor used in the 'original capture' process, vs. the range of the sensor used in the negative scanning process, as well as the intermediate representation, which in the DI case, at least some of the processes I've looked into, is a 'log' representation of the intensity values. Heck, the process may also represent intensity values as short floats rather than fixed integers...

This allows for greater 'processing' during the grading/compositing process, than using fixed integer size for each of R, G, and B. (or however the data is encoded.)

 

[Joe Calabrese]

Sharpness is perceived, and mostly can be contributed to contrast and the way edges are defined. It can also be added in post. Resolution and color depth cannot TRUELY be added.

Resolution is simply the pixel count. Color depth can be limited by compression codec and such. Like Ryan said (and I think everyone will agree) more information at the start will help things look better as you go through color correction and such. However, having all this information does not mean the picture is sharp. Your focus puller could have been completely drunk.

 

[Ryan Patrick O'Hara]

However, having all this information does not mean the picture is sharp. Your focus puller could have been completely drunk.

Hahaha, ss true as you are, I think we are assuming the variable of focus is not an issue.

 

[Joe Calabrese]

Hahaha, ss true as you are, I think we are assuming the variable of focus is not an issue.

Let's not worry about something that could be fixed in post. :thumbsup:

 

[h-munster]

Yes, you are right about most things, ...

Wait. You're saying that I'm not right about everything?

Dynamic range is also greater by a perceived effect of higher resolution. Meaning 52__ will appear to have increased dynamic range over 72__ kodak stock.

Higher resolution might extend the capture dynamic range ever so slghtly in the highlights and shadows, simply because there would be a few more tones available in the very top end and very bottom end.

I will say that I'm not sure there is an advantage of shooting 5-perf 65mm over 35mm when viewed on a 300x704 standard def DVD or 480x704 standard def DVD.

The only visual advantage may be that the contrast will be different. Higher resolution emulsions often give lower contrast (but not always). There might be a promotional advantage to being able to say, "This web series was shot on Imax!"

At that compression and tiny resolution, I don't know if it would be worth it.

Ah. That's the rub. One could have the same doubt about shooting hyperHD and outputting on standard HD. How important are the slight improvements?

 

[h-munster]

There is no doubt that if you compare 65mm film to 35mm film, both finished photo-chemically to there respective formats, that the 65mm film will be superior, all other variables being equal.

Well, what is "superior" can be subjective. People intentionally shoot super 8 and Pixel-Vision over high-end formats. Let's just say that higher resolution yields greater color depth.

Now you go down the food chain and do 65mm and 35mm, finished photo-chemically, but with the 65mm film being presented as a 35mm reduction. Even if the reduction process is done from the camera original negative, which I doubt anyone would do today, are you saying that the 65mm originated film will look better?

No. I am not taking any side, and I would not even say that higher resolution or greater color depth is "better."

A lot depends on the resolution and color depth inherent in the intermediate stage. If the intermediate stage has higher resolution and/or greater color depth than either or both of the originals, then a difference in "quality" should be perceptible.

Even if the intermediate stage is the same or less quality than the 35mm original, there will be a minute difference in the look of the two, partially due to the way the reduced 65mm grain meshes with the 35mm intermediate and because there is an ever so slight extension of the very extreme highlights and shadows with the higher res of 65mm.

And further down the food chain, doesn't thing get worse and worse, and the advantages of the uber high resolution origination kind of peak at a point, especially in regards to color depth?

I am sorry, but I do not understand this question.

In the real world, using real work flows, the advantages of higher resolution origination is an advantage only if the entire work flow supports that advantage.

For the most part, I agree, but the differences in look (grain meshing, slight highlight/shadow extension) mentioned above will still apply.

To me, that was one of the main findings of the Zacuto Shootout. The colorist made quit a point about the 35mm film being much easier to grade, and having much better color depth, but that it was not because of the HDDSLRs resolution, but from the HDDSLRs sh*tty origination codec.

Surely, there is more CC latitude with an uncompressed film scan compared to a compressed HDSLR video file.

 

[h-munster]

When going to 480x704 or less... whether one is shooting on 35mm, 65mm or 100,000mm really isn't a big deal.

Again, a lot depends on the other variables of the 480x704 stage. It's not just the resolution. If the bit depth is 16-bit, a difference should be apparent.

Of course, it would not be very practical to encode DVD footage at 16-bit.

 

[h-munster]

The higher resolution 'enhances' the representation of the image area. It does not change 'dynamic range' in the least.

It depends on whether one refers to the dynamic range ability of the capture (latitude) or whether one refers to the value range of the resulting image.

Higher resolution in the capture format will give more color depth in the resulting image, which means a few more in-between shades in the very top highlights and in the very bottom shadows, which should give a slightly extended dynamic range in the capture.

For example, if a gradient is made, and your digitizer only has 256 values, and the change from pixel to pixel is less than 1/256 of the range of the digitizer, you wil have banding whether you digitize to 1024 or 4096 pixels along a horizontal line...

Actually, if the number of possible values per pixel is constant, there will be less banding in the higher res scan. As resolution increases, so does color depth, thus more "bands" are created.

Naturally, if we take an already "banded" image, and just increase the resolution, the bands will still remain. But if we reduce the size of the banded image, the resolution increases (more pixels per area) as does the color depth, and, thus, the bands will start to blend and disappear.

The whole issue of 'viewing' with regard to 'resolution' and 'dynamic range', is intimately tied to 'viewing distance' as well...

Of course. As the viewing distance increases, so does the resolution (more pixels per viewing angle), and so does the color depth.

... it just so happens that 'silver film grain', aka 'noise' is more aesthetically pleasing than current 'digital noise' in the form of excessively 'enlarged' digital images...

This is a common misconception. Digital noise is true noise, whereas grain is not noise at all -- grain is crucial image information in what is essentially a one-bit imaging system.

Consider black and white film emulsion. It is often called "continuous tone." However, there are no actual gray shades, because the gray tones are actually various densities of black silver grain clumps against a white (or clear) field. This fact becomes obvious upon magnification:

On a microscopic level, there is either a black grain or a white/clear space at each point, therefore, what is known as "continuous tone" is actually a one-bit system. These tiny bits blend in the viewer's eye to form the different shades.

Surprisingly, a similar "blending" effect occurs event when the grain clumps are large enough to be perceived (as in figure "b" above).

Color film employs same method of blending adjacent microscopic areas within the viewer's eye, but, with color, things can be a little more complex.

For instance, most color emulsions feature miniscule color dye "clouds," which form after corresponding silver grains have been bleached away. Each dye cloud is one of three colors, usually cyan or magenta or yellow. The adjacent dye clouds blend within the viewer's eye, but they also overlap to form areas of separate, new, actual colors (which are not a result of blending).

Additionally, to enhance contrast, many motion picture print stocks are designed to retain some of the exposed, black silver grains, further complicating the color depth. The grain is definitely not considered to be "noise" in this instance.

Even with all of this complexity, the color-depth/resolution rule still applies -- the finer the dye clouds and their overlapping areas, the greater the number of colors possible from interblending.

When I was more 'in' with things at 'fuji' one of the methods that their sensors used to gain better highlight response was to have two sensor cells where one was less sensitive, and so, could respond without saturation to higher levels of 'light'. So 'Cell 1' would respond to low/mid light, "Cell 2" would respond to 'mid/high', and the result was 'increased dynamic range'... but I digress...

Panavision is about to start implementing a similar technique. There are other methods that change the sensitivity of individual cells dynamically, but such sensors are usually not great for cinematography.

I read a rumor that Red was going to attempt a way to get multi-sensitive cells.

To sum up, for me 'increased resolution' only gives better 'dynamic range' representation, when the underlying sensor/digitizer has high dynamic range.

The dynamic range increase that accompanies a rise in resolution is slight. However, the color depth can skyrocket from only a small boost in resolution.

 

[h-munster]

Sharpness is perceived, and mostly can be contributed to contrast and the way edges are defined.

I think that sharpness is actually a combination of resolution and of the contrast level of the "edge" details.

Resolution is simply the pixel count.

Well, with Bayer sensors and film emulsions, assessing the resolution is a little more complex.

Color depth can be limited by compression codec and such.

Only if one limits the resolution and bit depth.

Color depth is essentially a product of resolution and bit depth in most imaging systems. There is actually a formula that can be used to calculate color depth, given the resolution and bit depth. The formula can be found in the latter half of this page.

 

[Razz16mm]

With any fixed pattern digital sensor system, Bayer or not, your visible resolution will be at the very best around 85% of the pixel resolution unless you are willing to tolerate aliasing and moire in the finished image.

Also with digital systems, oversampling, ie higher resolution, greater color depth. greater dynamic range in the original will result in better image quality at finish resolution. You need 2x oversample to fully resolve to your finished format and differences will be noticeable up to 4x oversample. On TV for instance the difference in an SD program that originated in HD vs SD that was shot in SD is very obvious.
Greater dynamic range allows retention of detail in highlights and shadows that can be preserved in a lower dynamic range viewing format with appropriate grading and gamma curves.
Regardless of finish format, more information in the original gives more range for creative flexibility in post.

 

[Joe Calabrese]

I think that sharpness is actually a combination of resolution and of the contrast level of the "edge" details.

Resolution is mostly independent of sharpness. Yes, a high resolution image can look sharp when scaled down, but an image that is native 320x240 can still look sharp to our eyes. One person could say a picture looks too sharp, and another can say the same picture looks just ok. Its all relative.

 

[j1clark@ucsd.edu]

It depends on whether one refers to the dynamic range ability of the capture (latitude) or whether one refers to the value range of the resulting image.

Higher resolution in the capture format will give more color depth in the resulting image, which means a few more in-between shades in the very top highlights and in the very bottom shadows, which should give a slightly extended dynamic range in the capture.

I don't disagree that if the digitization has the capability to represent intensity differences, that capturing at a higher resolution will produce an image that has more intensity variation.

However, if the digitizer does not have the sensitivity to capture the variation, the resolution will have no effect on the resulting image.

An example would be to capture an image that has an 8 bit digitizer, then representing that as a histogram, using 16 bit values. One will have a histogram of 'spikes' corresponding to the 8 bit values, and 'zero' in between.

Then compare that to an image digitized to 16 bits, and one will see a histogram with values lying between the '8 bit' samples.

I'm thinking of the binary as a fixed point quantity lying between 0 and 1. where 0 = 0.0000, and 255 = .9999... I realize most people don't think about binary values in this way... but anyway...

So going to a 16 bit representation would yield values of 0.0000000 to 0.9999999

(I've not taken the time to work out the values exactly, but that's the just of viewing data between 0 and full scale input signal, then mapping that to a fractional number represented in fixed point integer form...)