Intel® Integrated Performance Primitives
Deliberate problems developing high-performance vision, signal, security, and storage applications.
6765 Discussions

ippiResizeSqrPixel_8u_C1R with x or y factor of 1.0 and SUPER

slabrozzi
Beginner
1,203 Views

For downsample I am trying to use ippiResizeSqrPixel_8u_C1R(IPP 6.1.2) and SUPER (because other scalealgorithms have very poor quality in downscaling). This works well except for a few cases. One specific case is where I need to resize horizontally or vertically but not in the other. For example, 640x240 to 320x240. When either the x or y factor is 1.0 and using the IPPI_INTER_SUPER ippiResizeSqrPixel_8u_C1R returns error ippStsResizeFactorErr.

0 Kudos
6 Replies
Chao_Y_Intel
Moderator
1,203 Views

Hello,

Thanks for posting it. We will check if this is a bug.

Thanks,

Chao

0 Kudos
Yuri_Tikhomirov__Int
New Contributor I
1,203 Views

Hi Slabrozzi

Please look at the documentation http://software.intel.com/sites/products/documentation/hpc/ipp/chm/ippiman.chm

Appendix B (Super Sampling): "Note that supersampling interpolation can be used only for xFactor<1 and yFactor<1"

I advise you to use the function ippiSuperSampling_8u_C1R for down-sampling. It has simple interface and faster than ippiResizeSqrPixel_8u_C1R for SS method.

ippiSuperSampling_8u_C1R (
const Ipp8u* pSrc, int srcStep, IppiSize srcRoiSize,
Ipp8u* pDst, int dstStep, IppiSize dstRoiSize,
Ipp8u* pBuffer)

Thanks,

Beg

0 Kudos
matthieu_darbois
New Contributor III
1,203 Views

Hi,

The documentation states that xFactor and yFactor should be less than one but given the algorithm used for super sampling, it should work also with a factor 1.0

I think that this could be considered as a bug since the algorithm itself accepts 1.0 as a factor.

Regards,

Matthieu

0 Kudos
slabrozzi
Beginner
1,203 Views
I've been ussing IPP for resize since IIP 2.0 and I did realize that 1.0 was not supported for super but was really hoping that this is something that could be considered for support. In many video applications often an interlaced video frame (say square pel at 640x480) is treated as a single field (640x240) and the desire is to have this reduced to 320x240. The quality differences between using super and non-super is extreme.
0 Kudos
PaulF_IntelCorp
Employee
1,203 Views

I've added a feature request to include the ability to use a 1.0 scale factor for these functions in the future. Please note that this is not a guarantee that such features will be added.

Paul

0 Kudos
PaulF_IntelCorp
Employee
1,203 Views
FYI, this feature has been implemented, but will not appear until the next major release.
0 Kudos
Reply