Hello,
I have compiled and run both the MPEG4 encoder sample for IA32/64 and PCA, under Linux/PC enviroment, with a little portting work.
The IA32 sample generates fairly good quanlity video, but for PCA
the PVOP encoding result is very poor. Looking at the code I find
that the IA32 sample doesn't use the IPP motion estimation functions
but implements its own version, using functions not available in the
PCA version of IPP. I also tried the MVFAST functiosn in IPP, but it
doesn't make any difference. My question is:
1. Is the motion estimation functions in IPP not as good as that
found in the IA32 sample implementation?
2. Do your guys have some guideline on improving inter-encoding
qunanlity of the PCA sample?
3. The documentation of the ME algorithms is not clear on some
function arguments, where to find more information about it?
I have compiled and run both the MPEG4 encoder sample for IA32/64 and PCA, under Linux/PC enviroment, with a little portting work.
The IA32 sample generates fairly good quanlity video, but for PCA
the PVOP encoding result is very poor. Looking at the code I find
that the IA32 sample doesn't use the IPP motion estimation functions
but implements its own version, using functions not available in the
PCA version of IPP. I also tried the MVFAST functiosn in IPP, but it
doesn't make any difference. My question is:
1. Is the motion estimation functions in IPP not as good as that
found in the IA32 sample implementation?
2. Do your guys have some guideline on improving inter-encoding
qunanlity of the PCA sample?
3. The documentation of the ME algorithms is not clear on some
function arguments, where to find more information about it?
連結已複製
1 回應
