YUV Formats

YUV formats fall into two distinct groups, the packed formats where Y, U (Cb) and V (Cr) samples are packed together into macropixels which are stored in a single array, and the planar formats where each component is stored as a separate array, the final image being a fusing of the three separate planes.

In the diagrams below, the numerical suffix attached to each Y, U or V sample indicates the sampling position across the image line, so, for example, V0 indicates the leftmost V sample and Yn indicates the Y sample at the (n+1)th pixel from the left.

Subsampling intervals in the horizontal and vertical directions may merit some explanation. The horizontal subsampling interval describes how frequently across a line a sample of that component is taken while the vertical interval describes on which lines samples are taken. For example, UYVY format has a horizontal subsampling period of 2 for both the U and V components indicating that U and V samples are taken for every second pixel across a line. Their vertical subsampling period is 1 indicating that U and V samples are taken on each line of the image.

For YVU9, though, the vertical subsampling interval is 4. This indicates that U and V samples are only taken on every fourth line of the original image. Since the horizontal sampling period is also 4, a single U and a single V sample are taken for each square block of 16 image pixels.

Microsoft has some MSDN pages defining YUV formats too. They seem to miss quite a few formats but may be interesting:

• 4:2:0 Video Pixel Formats
• 4:2:2 Video Pixel Formats

Also, if you are interested in YCrCb to RGB conversion, you may find this page helpful.

Packed YUV Formats

AYUV

This is a 4:4:4 YUV format with 8 bit samples for each component along with an 8 bit alpha blend value per pixel. Component ordering is A Y U V (as the name suggests).

UYVY (and Y422 and UYNV)

UYVY is probably the most popular of the various YUV 4:2:2 formats. It is output as the format of choice by the Radius Cinepak codec and is often the second choice of software MPEG codecs after YV12.

Effective bits per pixel : 16

Positive biHeight implies top-down imge (top line first)

IUYV

IUYV is basically the same as UYVY with the exception that the data is interlaced. Lines are ordered 0,2,4,....,1,3,5.... instead of 0,1,2,3,4,5,....

cyuv

This FOURCC, allegedly registered by Creative Labs, is essentially a duplicate of UYVY. The only difference is that the image is flipped vertically, the first u_int16 in the buffer representing the bottom line of the viewed image. Note that the FOURCC is comprised of lower case characters (so much for the upper case convention !)

Effective bits per pixel : 16

Positive biHeight implies bottom-up image (botton line first)

YUY2 (and YUNV and V422 and YUYV)

YUY2 is another in the family of YUV 4:2:2 formats and appears to be used by all the same codecs as UYVY.

Effective bits per pixel : 16

Positive biHeight implies top-down image (top line first)

There is a help page here which contains information on playing AVIs which include video stored in YUY2 format.

YVYU

Despite being a simple byte ordering change from YUY2 or UYVY, YVYU seems to be seen somewhat less often than the other two formats defined above.

Effective bits per pixel : 16

Positive biHeight implies top-down image (top line first)

Y41P

This YUV 4:1:1 format is registered as a PCI standard format. Mediamatics MPEG 1 engine is the only codec (other than a Brooktree internal one) that I know of that can generate it.

Effective bits per pixel : 12

Positive biHeight implies top-down image (top line first)

Y411

I was originally told that this was a duplicate of Y41P however it seems that this is not the case after all. Y411 is a packed YUV 4:1:1 format with a 6 pixel macroblock structure containing 4 pixels. Component packing order is:

Effective bits per pixel : 12

Positive biHeight implies top-down image (top line first)

IY41

IY41 is basically the same as Y41P with the exception that the data is interlaced. Lines are ordered 0,2,4,....,1,3,5.... instead of 0,1,2,3,4,5,....

Y211

I have yet to find anything that will output Y211 ! The format looks very much like the missing YUV 4:2:2 ordering but Y samples are only taken on every second pixel. Think of it as a half width 4:2:2 image and double the width on display.

Effective bits per pixel : 8

Positive biHeight implies top-down image (top line first)

Y41T

This format is identical to Y41P except for the fact that the least significant bit of each Y component forms a chromakey channel. If this bit is set, the YUV image pixel is displayed, if cleared, the pixel is transparent (and the underlying graphics pixel is shown).

Positive biHeight implies top-down image (top line first)

Y42T

This format is identical to UYVY except for the fact that the least significant bit of each Y component forms a chromakey channel. If this bit is set, the YUV image pixel is displayed, if cleared, the pixel is transparent (and the underlying graphics pixel is shown).

Positive biHeight implies top-down image (top line first)

CLJR

Cirrus Logic's format packs 4 pixel samples into a single u_int32 by sacrificing precision on each sample. Y samples are truncated to 5 bits each, U and V have 6 bits per sample.

Effective bits per pixel : 8

Positive biHeight implies top-down image (top line first)

IYU1

The IYU1 format is a 12 bit format used in mode 2 of the IEEE 1394 Digital Camera 1.04 spec ("1394-based Digital Camera Specification, Version 1.04, August 9, 1996", page 14.). The format, a duplicate of Y411, is YUV (4:1:1) according to the following pattern:

IYU2

The IYU2 format is a 24 bit format used in mode 0 of the IEEE 1394 Digital Camera 1.04 spec (ibid.) The format is YUV (4:4:4) according to the following pattern:

YUVP

This is another format similar to YUY2 and it's aliases. The difference here is that each Y, U and V samples is 10 bits rather than 8. I am still waiting to hear how the samples are packed - is a macropixel just 5 bytes long with all the samples packed together or is there more to it than this?

V210

AJA Video Systems have implemented this Quicktime format for Windows. It is a 10 bit per component, YCrCb 4:2:2 format in which samples for 5 pixels are packed into 4 4-byte little endian words. Rather than repeat the details here, I suggest looking at the original definition on the Quicktime web site.

Planar YUV Formats

YVU9

This format dates back to the days of the ActionMedia II adapter and comprises an NxN plane of Y samples, 8 bits each, followed by (N/4)x(N/4) V and U planes.

Positive biHeight implies top-down image (top line first)

ATI has a codec supporting this format that you can download from here.

IF09

A derivative of YVU9, IF09 contains the basic 3 planes for Y, V and U followed by an additional (N/4)x(N/4) plane of "skip blocks". This final plane forms a basic delta encoding scheme which can be used by a displayer to decide which pixels in the image are unchanged from the previous displayed frame. The strange number of bits per pixel listed for the format results from the fact that an NxN image is described using N²+3(N/4)² bytes.

This format is generated by Intel's Indeo codecs though users should beware - the original 32 bit Indeo 3.2 shipped with Windows 95 and the beta levels of Indeo 4.1 contain bugs which cause them to generate protection faults when using IF09. Fixed versions of these codecs are available from Intel.

Positive biHeight implies top-down image (top line first)

Delta plane definition

To be completed...

YV12

This is the format of choice for many software MPEG codecs. It comprises an NxM Y plane followed by (N/2)x(M/2) V and U planes.

Positive biHeight implies top-down image (top line first)

ATI has a codec supporting this format that you can download from here.

YV16

This format is basically a version of YV12 with higher chroma resolution. It comprises an NxM Y plane followed by (N/2)xM U and V planes.

IYUV and I420

These formats are identical to YV12 except that the U and V plane order is reversed. They comprise an NxN Y plane followed by (N/2)x(N/2) U and V planes. Full marks to Intel for registering the same format twice and full marks to Microsoft for not picking up on this and rejecting the second registration.

Positive biHeight implies top-down image (top line first)

CLPL

This format introduces an extra level of indirection in the process of accessing YUV pixels in the surface. Locking the DirectDraw or DCI CLPL surface returns a pointer which itself points to three other pointers. These pointers respectively point to an NxN Y plane, an (N/2)x(N/2) U plane and an (N/2)x(N/2) V plane. The Y plane pointer retrieved is (allegedly) valid even when the surface is subsequently unlocked but the U and V pointers can only be used with a lock held (as you should be doing anyway if adhereing to the DirectDraw/DCI spec).

Positive biHeight implies top-down image (top line first)

Y800

This format contains only a single, 8 bit Y plane for monochrome images. Apparent duplicate FOURCCs are "Y8" and "GREY".

NV12

YUV 4:2:0 image with a plane of 8 bit Y samples followed by an interleaved U/V plane containing 8 bit 2x2 subsampled colour difference samples.

Microsoft defines this format as follows:

 "A format in which all Y samples are found first in memory as an array of unsigned char with an even number of lines (possibly with a larger stride for memory alignment), followed immediately by an array of unsigned char containing interleaved Cb and Cr samples (such that if addressed as a little-endian WORD type, Cb would be in the LSBs and Cr would be in the MSBs) with the same total stride as the Y samples. This is the preferred 4:2:0 pixel format."

NV21

YUV 4:2:0 image with a plane of 8 bit Y samples followed by an interleaved V/U plane containing 8 bit 2x2 subsampled chroma samples. The same as NV12 except the interleave order of U and V is reversed.

Microsoft defines this format as follows:

 "The same as NV12, except that Cb and Cr samples are swapped so that the chroma array of unsigned char would have Cr followed by Cb for each sample (such that if addressed as a little-endian WORD type, Cr would be in the LSBs and Cb would be in the MSBs)."

IMC1

Similar to YV12, this format comprises an NxN Y plane followed by (N/2)x(N/2) U and V planes. The U and V planes have the same stride as the Y plane and are restricted to start on 16 line boundaries.

Microsoft defines this format as follows:

 "The same as YV12, except that the stride of the Cb and Cr planes is the same as the stride in the Y plane. The Cb and Cr planes are also restricted to fall on memory boundaries that are a multiple of 16 lines (a restriction that has no effect on usage for the standard formats, since the standards all use 16×16 macroblocks)."

IMC2

Similar to YV12, this format comprises an NxN Y plane followed by "rectangularly adjacent" (N/2)x(N/2) U and V planes. Lines of U and V pixels are interleaved at half stride boundaries below the Y plane.

Microsoft defines this format as follows:

 "The same as IMC1, except that Cb and Cr lines are interleaved at half-stride boundaries. In other words, each full-stride line in the chrominance area starts with a line of Cr, followed by a line of Cb that starts at the next half-stride boundary. (This is a more address-space-efficient format than IMC1, cutting the chrominance address space in half, and thus cutting the total address space by 25%.) This runs a close second in preference relative to NV12, but NV12 appears to be more popular."

IMC3

The same as IMC1 except for swapping the U and V order.

IMC4

The same as IMC2 except for swapping the U and V order.