What is 4:2:2, 4:4:4, etc...

[Paul Arrojado]

..and how does it affects the image/film quality?!? and/or the output?!

 

[Yousuf Abbasi]

subsampling technique which describes how chroma is sampled relative to 4 parts of luma. So, 4:2:2 samples half as much chroma info as luma. 4:2:0 is even more subsampling where the sampling alternates from sampling half as much chroma per unit luma, to zero chroma info per unit luma. 4:4:4 is true RGB, meaning no color subsampling has taken place. The output is essentially you have better color information! The reason more luma is generally sampled relative to chroma is due to belief that humans percieve brightness better than color. It's all just a way to get a decent image across using least amount of bandwith.

 

[Dan Hudgins]

Noise

Noise

Because of sensor noise, I would think that even 3 chip RGB cameras are using area chroma filters that are needed for Bayer filter cameras to reduce chroma noise.

Becasue of the heavy digital filters now used before compression probably in all modern cameras, there is NO camera that records RGB sensor data even in 3 chip cameras, what you see is a processed image in which the chroma information has been area filtered (with edge detection) and gambit remaped using 3D LUT.

Even film scans are cleaned up using chroma and other noise filters. We have a film scanner that shoots true RAW RGB images, and without processing the results do not look like the scans you see being used, the sharpness of the scans shows color grain in the film that would normally be blured in film to film printing.

So EVERYTHING gets processed before compression or recording, 4:4:4 is a myth for the most part now, it may define a way of saving data but the data being saved has undergone chroma averaging in such a way that there is no single pixel chroma detail in the normal sence of the idea, like in a drum scanner. If you did have full color for each pixel the images would have heavy chroma noise in them because the sensors are not as good as they look, the have much higher native noise in the color, more so with Bayer sensors because of the contrast boost required to saturate the chroma from the weak colored filters.

People should stop asking about issues like 4:4:4 vs. this or that, those are not how the cameras work inside even if they are marketed under buzz words.

Its better to look at the images each camera makes and see if it fits your needs, specs are so far from what is going on with the image data to be for the most part useless in predicting at what stage in the image processing the image becomes useless or usefull...

 

[Alan Skinner]

Dan, is this a typo on your signature?

"intergrated color correction"

Should it be "integrated"?

Alan

 

[Stacey Spears]

So, 4:2:2 samples half as much chroma info as luma. 4:2:0 is even more subsampling where the sampling alternates from sampling half as much chroma per unit luma, to zero chroma info per unit luma. 4:4:4 is true RGB

4:4:4 - Chroma and Luma have the same resolution. There is a 4:4:4 Y'CbCr.

4:2:2 - Horizontal chroma resolution is half of Luma.

4:2:0 (Progressive) - Horizontal and vertical chroma resolution is half of Luma.

4:2:0 (Interlaced) - Horizontal and vertical chroma resolution is half of Luma.

The images are from an article I wrote way back when. http://www.hometheaterhifi.com/the-...or-and-the-420-interlaced-chroma-problem.html

 

[Dan Hudgins]

Thanks

Thanks

Dan,
Should it be "integrated"?
Alan

Thanks for pointing that out, I edited it.

==

The ways the color information are encoded does not tell you how the chroma data was filtered before encoding. As we know Bayer filter cameras can show less detail in the Blue and Red than Green, so thats why some high end HD cameras use Quad-HD chips to get four sensor pixels for each output pixel. You would think that Quad-HD would give you chroma for each pixel in 4:4:4, but infact the internal processing in the camera blends the chroma through low pass and edge filters before the 4:4:4 encoding. If you make a true RGB 4:4:4 three chip camera you would get high noise in the chroma because if you build the camera like a RGB single shot three plate view camera with strong filters you get lower EI/ISO in the LUMA.

So those encoding systems do not indicate the chroma resoultion of something shoot through a lens on any given pixel and its neighbors. It is true that if you save the chroma at a lower bandwidth you cannot put the edges back where they were, so there is some advantage to higher bandwidth encoding, but the cameras do not work as RGB like the old Technicolor film camera shooting three monochrome images with three chips and saving that data, internal to the modern camera are complex filters that try to reduce the sensor noise to get a higher EI/ISO, these filters are proprietary and the mfg do not want to talk about the single pixel chroma accuracy of their cameras, i.e. a single red pixel on blue background, because the filters used to reduce aliasing, chroma moire, and noise all work against fine chroma details, so with high bandwidth recording you just get a better recording of a blurred chroma image, hence the trend to record the chroma at lower bandwidths since it ends up about the same to your eye.

 

[Yousuf Abbasi]

The images are from an article I wrote way back when. http://www.hometheaterhifi.com/the-...or-and-the-420-interlaced-chroma-problem.html

Pretty comprehensive informative article. I glanced through it and will read it on the subway. :thumbsup: