4:4:4 - 4:1:1 - 4:2:1 - Explaination please ?
- Thread starter softfx
- Start date
Hello softfx:
There are several ways in which the color of a pixel in a video image can be encoded. For example, one is RGB, where each pixel has 3 values, one for Red, one for Green and one for Blue. Combined, this three values give the actual color of the pixel(kinda like water color).
Most video codecs however use a system called YUV in which Y contains the luminosity of the pixel, and U and V contain the color. To understand what Y means, if you took an image incoded in YUV, and removed the U and V, you would get a black and white (grayscale) image.
The 4:4:4 or 4:1:1(..etc) notations simply state, for a particular image, how much information is recorded for each component. For example, with 4:4:4, all the information is recorded. That is, for any pixel in the video image, there is a distinct Y, U and V value. In 4:1:1, there is luminosity(Y) recorded for every pixel. However, U and V are only recorded every 3 pixels. This reduces the amount of information so it fits on a smaller format(such as tape), but of course it brings some problems in image quality.
There is a good reason why this is done in this manner. The human eye is more sensitive to changes in luminosity than color, so luminosity basically judges how 'sharp' things look. This is why you can get away with not recording full color, and still have it look good. The problem that 4:1:1 or any other level of decimation brings is when you want to do something to the footage such as greenscreen work or color correction. It really helps to have as much information there to work with.
The contrast ratio in projectors simply dictates the maximum difference in brightness between light and dark in the image you see. Essentially you want this to be as large as possible so you can show more dynamic range from your footage. For example, if you watch footage that was captured at a wide dynamic range such as film, on a projector with a low contrast ratio, it will look more like video, because the projector cannot reproduce all the tonality of the original material.
Hope this helps.
There are several ways in which the color of a pixel in a video image can be encoded. For example, one is RGB, where each pixel has 3 values, one for Red, one for Green and one for Blue. Combined, this three values give the actual color of the pixel(kinda like water color).
Most video codecs however use a system called YUV in which Y contains the luminosity of the pixel, and U and V contain the color. To understand what Y means, if you took an image incoded in YUV, and removed the U and V, you would get a black and white (grayscale) image.
The 4:4:4 or 4:1:1(..etc) notations simply state, for a particular image, how much information is recorded for each component. For example, with 4:4:4, all the information is recorded. That is, for any pixel in the video image, there is a distinct Y, U and V value. In 4:1:1, there is luminosity(Y) recorded for every pixel. However, U and V are only recorded every 3 pixels. This reduces the amount of information so it fits on a smaller format(such as tape), but of course it brings some problems in image quality.
There is a good reason why this is done in this manner. The human eye is more sensitive to changes in luminosity than color, so luminosity basically judges how 'sharp' things look. This is why you can get away with not recording full color, and still have it look good. The problem that 4:1:1 or any other level of decimation brings is when you want to do something to the footage such as greenscreen work or color correction. It really helps to have as much information there to work with.
The contrast ratio in projectors simply dictates the maximum difference in brightness between light and dark in the image you see. Essentially you want this to be as large as possible so you can show more dynamic range from your footage. For example, if you watch footage that was captured at a wide dynamic range such as film, on a projector with a low contrast ratio, it will look more like video, because the projector cannot reproduce all the tonality of the original material.
Hope this helps.
Pertier:
If U and V contains color, what colors? How would U be different than V? Does 4:2:0 mean color is missing in the V?
I had the (probably mistaken assumption) that YUV was some version of RGB.
Is UV related to the UV mapping coordinates of a 3D object?
I suppose I could Google the subject... but I don't feel like digging right now....
Thanks!
If U and V contains color, what colors? How would U be different than V? Does 4:2:0 mean color is missing in the V?
I had the (probably mistaken assumption) that YUV was some version of RGB.
Is UV related to the UV mapping coordinates of a 3D object?
I suppose I could Google the subject... but I don't feel like digging right now....
Thanks!
khmuse said:
AAieeee!
My head hurts.
I decided to become unlazy and google YUV and here's a good site:
http://en.wikipedia.org/wiki/YUV
The color graphic explains a lot to me.
Thanks
The basic concept to understand is that all the necessary information to obtain the color for the pixel is in the YUV values. U and V contain information that when combined with the grayscale image(described by Y) result in a color image.
4:2:0 is not exactly as it looks. At first glance it sounds like V is not sampled at all but this is not the case. A more intuitive notion would be: on the first line of the image, it's 4:2:0, on the second line it is 4:0:2, on the third line it is 4:2:0 again, and so on.
The 4:2:0 notation arises just from the fact that there is no way to describe the switching back and forth of U and V samping with the X:X:X notation.
Hope this helps,
Juan
4:2:0 is not exactly as it looks. At first glance it sounds like V is not sampled at all but this is not the case. A more intuitive notion would be: on the first line of the image, it's 4:2:0, on the second line it is 4:0:2, on the third line it is 4:2:0 again, and so on.
The 4:2:0 notation arises just from the fact that there is no way to describe the switching back and forth of U and V samping with the X:X:X notation.
Hope this helps,
Juan
mikkowilson
Steadi-Guru
hehe, regarding 4:2:0...
4:4:4 is full bandwidth. (actually in this state RGB not YUV is normally used.. but 4:4:4 is used to describe beacuse it fits in well with the others) of every 4 pixels they all contian all color information.
This is what VGA for example works at, or 'uncompressed' video
4:2:2 is 50% less color infomation. This is the 'highest' YUV format normally used. For every 4 pixels, the all of course carry lume, but only 2 (every other one) caries U and V respectivly.
DigiBeta, DVCPRO50 and Betacam (and all other analog 'component' formats) are all 4:2:2
4:1:1 is of course only 25% of the color information. Again for every 4 pixels you get all 4 samples of Y, but then only every 4th is U and every 4th is V.
DVCPRO, DV/MiniDV (NTSC) and DVCAM (NTSC) are all 4:1:1
now here's the goofy one:
4:2:0 ...is ACTAULLY "4:2:0 - 4:0:2"
but we only say 4:2:0 because the 4:0:2 is implied. This mean that of the first group of 4 pixels you gat all the Y (as normal) and 2 of the U, but NO V. of the next goup of 4, you again get all 4 Y, but this time NO U and then 2 of V. The final end result is actualy the same amount of information as 4:1:1, but just sampled slightly differently. The down side is that the colro is more acurate in little "chunks" in the groups of 2 samples, tehn tehre's a bgi space. the result is an EVEN worse average color resolution at most points, resulting in a very blocky chroma key if you have ot use 4:2:0
4:2:0 is only used in PAL countries, for DV/MiniDV and DVCAM. - DVCPRO is always 4:1:1, even in PAL.
I hope this clears up a little bit.
(what's really clever is how all this, all 3 dimensions of color, can be packed into a simgle dimensional analog signal.. but that's another [related] topic!)
- Mikko
4:4:4 is full bandwidth. (actually in this state RGB not YUV is normally used.. but 4:4:4 is used to describe beacuse it fits in well with the others) of every 4 pixels they all contian all color information.
This is what VGA for example works at, or 'uncompressed' video
4:2:2 is 50% less color infomation. This is the 'highest' YUV format normally used. For every 4 pixels, the all of course carry lume, but only 2 (every other one) caries U and V respectivly.
DigiBeta, DVCPRO50 and Betacam (and all other analog 'component' formats) are all 4:2:2
4:1:1 is of course only 25% of the color information. Again for every 4 pixels you get all 4 samples of Y, but then only every 4th is U and every 4th is V.
DVCPRO, DV/MiniDV (NTSC) and DVCAM (NTSC) are all 4:1:1
now here's the goofy one:
4:2:0 ...is ACTAULLY "4:2:0 - 4:0:2"
but we only say 4:2:0 because the 4:0:2 is implied. This mean that of the first group of 4 pixels you gat all the Y (as normal) and 2 of the U, but NO V. of the next goup of 4, you again get all 4 Y, but this time NO U and then 2 of V. The final end result is actualy the same amount of information as 4:1:1, but just sampled slightly differently. The down side is that the colro is more acurate in little "chunks" in the groups of 2 samples, tehn tehre's a bgi space. the result is an EVEN worse average color resolution at most points, resulting in a very blocky chroma key if you have ot use 4:2:0
4:2:0 is only used in PAL countries, for DV/MiniDV and DVCAM. - DVCPRO is always 4:1:1, even in PAL.
I hope this clears up a little bit.
(what's really clever is how all this, all 3 dimensions of color, can be packed into a simgle dimensional analog signal.. but that's another [related] topic!)
- Mikko
The xxx notation relates to a line based sampling raster where the luminance component is being generated out of the sampled RGB values at a standarized number per line (e.g.720 for DV) and the chroma components are generated in a given ratio as compared to the number of luma components like 1/4 (720/4)for a 4:1:1 raster. The "1:1" or "2:0" does not relate to U or V but to the chroma samples in succesive lines. So 4:2:0 means that chroma is being generated once at 2/4 (360 samples)of the luma value in one line and 0 times in the next line. Within the 4:2:0 notition there are still some variants depending on how the two chroma samples are spread between the lines (Difference between PAL and DVD )
excellent information. I've been trying to understand that clearly myself for awhile.
Now how does this relate to picture quality when you add one of those USB things to the DVX so that you can capture the DVX footage at 4:4:4. on an external hard-drive
Does that mean the video is edited at a very high quality compared to the other ratios, and what would that mean for file sizes?
Thanks
Sol
Now how does this relate to picture quality when you add one of those USB things to the DVX so that you can capture the DVX footage at 4:4:4. on an external hard-drive
Does that mean the video is edited at a very high quality compared to the other ratios, and what would that mean for file sizes?
Thanks
Sol
FaceTheSpencer
Active member
This bypasses the DV compression and allows you to pull color information off the chips for each pixel (making the 4:4:4 image). You'll also notice, all discussion had basically estimated 100gb=1hr. MUCH larger file sizes.
I am sure Juan will have much more to say, being that the Andromeda is his baby.
I am sure Juan will have much more to say, being that the Andromeda is his baby.
Sol,
The Andromeda device adds a USB plug to your DVX such that you can connect it to a computer for recording uncompressed 4:4:4. Our device takes the signals right after they are captured by the sensors, so you get several advatanges over the compressed DV output:
1.Full color sampling(4:4:4). All the information from the 3 CCD's is recorded, as opposed to the 4:1:1 scheme which discards about 50% of the information.
2.No compression. The signal is competely uncompressed and thus has no compression artifacts at all.
3.The sensors on the DVX are actually 770x492, so you not only get more pixels in SD, but using the pixel shift between the sensors our system outputs images larger than 770x492.
4.More dynamic range. Our tests indicate that the Andromeda output shows about 2.5 stops of dynamic range over the DV output.
5.More color precision. Because the DVX actually works internally at 12-bits/channel(36-bit color total), our system supports recording at more than the 8-bits supported by DV, DVCPROHD and HDV.
You can find all this and more information at www.reel-stream.com, including sample images.
This system is not meant to be a perfect solution for all situations. It offers some unique advantages which no other system in this price range can offer, but there are also other things to consider such as needing to have a computer/laptop connected for recording, etc.
Let me know if you have any other questions.
Cheers,
Juan
The Andromeda device adds a USB plug to your DVX such that you can connect it to a computer for recording uncompressed 4:4:4. Our device takes the signals right after they are captured by the sensors, so you get several advatanges over the compressed DV output:
1.Full color sampling(4:4:4). All the information from the 3 CCD's is recorded, as opposed to the 4:1:1 scheme which discards about 50% of the information.
2.No compression. The signal is competely uncompressed and thus has no compression artifacts at all.
3.The sensors on the DVX are actually 770x492, so you not only get more pixels in SD, but using the pixel shift between the sensors our system outputs images larger than 770x492.
4.More dynamic range. Our tests indicate that the Andromeda output shows about 2.5 stops of dynamic range over the DV output.
5.More color precision. Because the DVX actually works internally at 12-bits/channel(36-bit color total), our system supports recording at more than the 8-bits supported by DV, DVCPROHD and HDV.
You can find all this and more information at www.reel-stream.com, including sample images.
This system is not meant to be a perfect solution for all situations. It offers some unique advantages which no other system in this price range can offer, but there are also other things to consider such as needing to have a computer/laptop connected for recording, etc.
Let me know if you have any other questions.
Cheers,
Juan
Graeme_Nattress
Red Team
Here's an explanation with pictures:
http://www.nattress.com/Chroma_Investigation/chromasampling.htm
Enjoy.
Graeme
http://www.nattress.com/Chroma_Investigation/chromasampling.htm
Enjoy.
Graeme
mikkowilson
Steadi-Guru
AWSOME link Graeme!
NOT only a good visual explanation of color sampling, but also some really interesting tests. - That resulted in prety much exactly what I'd expect.
Maybe that should be added to the links forum.
- Mikko
NOT only a good visual explanation of color sampling, but also some really interesting tests. - That resulted in prety much exactly what I'd expect.
Maybe that should be added to the links forum.
- Mikko
J.R. Hudson
Spirit Animal
softfx said:
Who cares. :thumbsup: :laugh: :engel017: :evil:
mikkowilson
Steadi-Guru
Techies care... :thumbsup: