Intel® Integrated Performance Primitives
Community support and discussions relating to developing high-performance vision, signal, security, and storage applications.
Announcements
The Intel sign-in experience is changing in February to support enhanced security controls. If you sign in, click here for more information.
6656 Discussions

Some error in ipp reference manual volume 2, video coding...

abuatnext04
Beginner
160 Views
Hi guys...

i have found some small error in documentation,
1) figure 16-29 DV-encoding Pipeline :
The same value of type block (2) is used to 8x8 and 2x4x8 mode...
2) DTC2x4x8Frw : the KC defintion is write as cos(pi*u(2z+1)/8)cos(pi*u(2z+1)/16) but the correct definition is cos(pi*u(2z+1)/8)cos(pi*h(2x+1)/16)

an other point is about compatibilities between ipp function and other. An example, is that i need to build a transcoder DV so, i use ipp function to decode a video segment and an other librairie to perform recoding. But the exact precision and interval of data in each ipp function is not explicit. So, if someone want to use ipp decoding and unquantize dv dc and ac coefficient, it's require an uneasy adaptation if we don't know the exact output data range...

thanks, i hope it could be in future version
0 Kudos
1 Reply
Vladimir_Dudnik
Employee
160 Views

Hi,

thanks for pointing us to that, you are right, there are some mistakes in documentation which will be fixed in future versions. Remember, you can submit your suggestions and issue reports to IPP technical support through Intel Premier Support.

1) block_type should be 1
2) you are right, coefficients should be as you mentioned

Regards,
Vladimir

Reply