So 1 pixel is 3x16 = 48 bit. One block is addressed leaving 1 peace of 16 bit,1 word, left for other use like alpha channel.
George
I use softwares which use 4*32 bit depht by default with resolution up to 30000 x 30000 pixels (I never had to use more but it is possible).
Some are even working in double precision (2 words) for computation.
And there are no problem of memory or amount of adress avalaible.
What software is that?
George
3d rendering softwares. And they have to store lot of other images (generally never less than 4k 8bit, 16bit or 32 bits) for computation (textures, environments, etc âŚ).
They âall in oneâ do what we do when taking a shot :
they illuminate the scene, take a shot and process the shot (the only stage which does not exist is demosaicing of course - even AI denoising stage exists).
I donât know what you mean with 3d software. The 3d software I think of are technical drawing programs like Autocad. They are vector based.
George
Autocad is more CAD oriented.
I use softwares for making films, advertising, etc ⌠So for making top images.
Most rendering softwares are plugging (or standalone). Even if 3D application have their rendering engine, plugin solution are completly integrated and transparent so as easy to use than proprietary renderers.
Look at âoctaneâ for example (I like this one because it is a âfull spectral rendererâ and not a ârgb rendererâ).
Works in 96 + alpha bit depth (and can save 96 bit + alpha images if needed). Can output (so store when computing) tens of different buffers in addition to main buffer. Applies post process to choose any color space, to emulate film (as in color rendering in photolab with filmpack), to choose your response curve, apply gamma curve, etc ⌠etc âŚ
And all this on very very very high resolutions if needed (way higher than any camera can do).
All I talk about is pixel based.
But Autocad does not work in vectors when itâs about rendering an image (it can do that too).
As EVERY renderer, it is a pixel based (and more) process. It creates pixel images with a pixel processing (and more) solution.
it (Octane) even allows deeppixel rendering which is huge compared to 96 bit images (it contains a huge depht buffer which allow to mix images together knowing what is front or back between those two (or more) images in compositing (it has to be very precise).
This is for heavy pipelines but works on any actual workstation.
I donât know anything about that business.
George
I think photographic software devellopers should look more closely what exist in this âbusinessâ and how itâs done (3d rendering, compositing).
Photography is the poor child in many ways when itâs about processing digital images.
Color Gamut has more to do with the range of colors represented than computer memory or the bit-depth of an image.
This example shows how many more colors AdobeRGB can represent compared to iSRGB. Presumably, DxOâs Wide Gamet represents even more colors than AdobeRGB.
The value of the greater range of colors is not so much in representing the colors in the original image as it is to provide latitude when editing the image to avoid âclippingâ colors.
Eventually, any gamut will have to be mapped to the output device. And some colors may not be faithfully reproduced in a smaller gamut such as iSRGB (many computer monitors) or the available gamut in a printer. But itâs always better to have a really large working gamut.
See here for DxOâs explanation of their WG:
1 Like
As far as i understood dxoâs reasons to create dxo WGC:
1 feature demand of many to support prophoto.
2 due the fact that dxo deepprime(xd) can restore more out the deepshadows so we can dig for more detail in the deep shadows, so we need more actual color differences in the lower part of the horseshoe. The saturated and darker colors who mostly excist in shadows.
And there is a small bit of more headroom in the brightest parts which was also a thing of people highlight recovery was not on par with say lightroom according to the feature request.
So for images which have a lot of clipping in shadows in legacy mode gets compressed colors in shadow parts.(they are rendered inside AdobeRGB colorspace by the protect saturated color algorithm) you could use WGC to raise shadows and get more detail out the shadows because there are not (less)compressed in the proces of creating the pixels in the working colorspace.
Not more colors but a wider range. The amount is defined by the bit depth. With a constat bitdepth and a wider gamut the step size is getting larger.
No, this is due to the bit depth mostly.
But be careful that more is not always better.
George
with AI you can invent, not restore
Yes, but with a given bit depht you can describe a given number of colors. No more, no less.
If you have more colors to describe (wider gamut), you have less number left to describe âin betweenâ colors than with a narrower gamut (since you have the same amount of numbers for describing more colors).
But anyway, 65 536 values per channel (16 bits) is probably enough to describe any existing visible subtlety in any color space.
1 Like
Thatâs what my grandmother used to say to me when I asked her for sweets.
A wise grandmother. 
George
1 Like
If an image is mapped into a small (say sRGB) gamut and your editing drives a color out of gamut, it is clipped and cannot be represented no matter how many bits are used. A larger gamut may not clip that color.
65,536 is 8 bits. 16 bits is 16.7 million. Per RGB channel. Plus luminance. The number of color possibilities if coded as 16 bits likey beyond the ability of human vision. Hence a larger color gamut is going to allow more colors to the limit of our vision.
Numbers are not colors : a given number does not represent the same color in different color spaces.
So if you have a finite number (given by the bit depht) to describe colors you will be able to describe more âintermediateâ colors in a small color space than in a wider color space, since you can describe the same amount of colors in both color spaces, but one is bigger than the other.
1 Like