Hello, I'm a master of Tongji University. Now I'm trying to make a
MPEG4 SP encoder based on BF533, and I use OmmVision OV6630 CIF CMOS
sensor to transfer original YCbCr data into BF533 via PPI. but the
sensor's output format is 4:2:2, just like belows:
...Cb Y Cr Y Cb Y Cr Y Cb...

...Cb Y Cr Y Cb Y Cr Y Cb...

...Cb Y Cr Y Cb Y Cr Y Cb...

...Cb Y Cr Y Cb Y Cr Y Cb...
4:2:2 format

While MPEG4 encoder input should be 4:2:0 format, so I have to
change 4:2:2 into 4:2:0.
...Y Y Y Y ...
...Cb Cr Cb...
...Y Y Y Y
...Cb Cr Cb...
...Y Y Y Y...
...Cb Cr Cb...
...Y Y Y Y...
4:2:2 format
I don't think I can just abandon half of the CbCr components simply
(for that will result in frequency interleaving).
Formally, a lowpass filter should be used to decrease CbCr band
width, then extract half of the CbCr components. But the mpeg4 encoder
should be real-time, indicating that I have to change 4;2:2 into 4:2:0
very fast,efficiently.So is there anyone who once solve this problem
perfectly? or anyone can find a better way? Help me, please! Start a New Thread