reset password
Author Message
lhairap
Posts: 11
Posted 13:51 Nov 11, 2015 |

I just need clarification on part F2(Subsampling) and I3(Supersampling).

In F2, for subsampling we find Cb and Cr, using 4:2:0, which is the average of 4 pixels. Then each one of those 4 pixels are getting that average value as their Cb and Cr. (Am I right?) I'm assuming after this step each pixel has Cb and Cr. 

If the above explanation is correct, I don't understand what should be done for I3 first part: "Supersample Cb and Cr so that each pixel has Cb and Cr."

vsluong4
Posts: 87
Posted 14:25 Nov 11, 2015 |

The idea behind subsampling is to decrease the filesize.  When we use 4:2:0 chroma, we're taking 4 Cr/Cb values and getting 1, a savings of 4x.  It would be pointless to do subsampling, get the representative value and store it 4 times, so we only need to store it once.

 

The problem is that when we convert back to RGB color space, we're short about 3/4 of the Cr/Cb values, so we supersample to get the correct number back.

lhairap
Posts: 11
Posted 14:49 Nov 11, 2015 |

Right, so after supersampling each one of 4 pixels will get the average value of Cb and Cr.

Thank you.