Doug Kerr
Well-known member
In a JPEG file, the pixel colors are not described in terms of their RGB coordinates, but rather in a luma (quasi-luminance), chroma (quasi-chrominance) form (YCC). (The chroma value is two-dimensional in the mathematical sense, consisting of two numerical values.)
There is a luma value for every pixel in the image, but not necessarily a chroma value for every pixel. When there is not one for every pixel, this is described as "chroma subsampling" (sometimes, inaccurately, as "chrominance subsampling").
In effect, the chroma information is at a lower resolution than the luma information. This is consistent with the fact that in human vision, the eye's resolution for luminance is greater than for chrominance.
There are identified in a peculiar notation, with the following choices being common. In the list, the first item is the designation, the second is the ratio of chroma values to pixels (stated as we see since the pattern is organized over clusters of 4 pixels:
The choice affects the size of the JPEG file for a given image. For example, for one image, the JPEG file size for three chroma subsampling schemes was as follows:
So we see that (at least for this image) the difference is not gigantic.
What do we know from experience about the effect of choosing one of these on the perceived "quality" of the resulting image?
Thanks.
Best regards,
Doug
There is a luma value for every pixel in the image, but not necessarily a chroma value for every pixel. When there is not one for every pixel, this is described as "chroma subsampling" (sometimes, inaccurately, as "chrominance subsampling").
In effect, the chroma information is at a lower resolution than the luma information. This is consistent with the fact that in human vision, the eye's resolution for luminance is greater than for chrominance.
There are identified in a peculiar notation, with the following choices being common. In the list, the first item is the designation, the second is the ratio of chroma values to pixels (stated as we see since the pattern is organized over clusters of 4 pixels:
4:1:0 1/8 (rarely used, included just for completeness)
4:1.1 2/8
4:2:0 2/8 (in a different way)
4:2:2 4/8
4:4:4 8/8 (not "subsampling", actually)
4:1.1 2/8
4:2:0 2/8 (in a different way)
4:2:2 4/8
4:4:4 8/8 (not "subsampling", actually)
The choice affects the size of the JPEG file for a given image. For example, for one image, the JPEG file size for three chroma subsampling schemes was as follows:
4:1:1 116 kB
4:2:2 129 kB
4:4:4 147 kB
4:2:2 129 kB
4:4:4 147 kB
So we see that (at least for this image) the difference is not gigantic.
What do we know from experience about the effect of choosing one of these on the perceived "quality" of the resulting image?
Thanks.
Best regards,
Doug