Intel® Integrated Performance Primitives

beginners questions

I'm drowning in IPP and MKL documentation
I have been using nsp up till now but now consider changing to IPP or MKL.
It would be very helpful for me if I can get some suggestions on porting.
I'll then would have a starting point.
Myapplication works with lots of large (512*512 up tp 8k*8k) images and fft's,but only one in memory at a time.
I have used the strategy of allocating each row for it self, both image and fft.
That means that I have to build coulmn vectors myself for the vertical FFTwhich slows things down I'll guess.

non-pic linking model possible?


I noticed that the intel's libipp*merged.a are compiled in PIC format (like -fPIC flag of gcc). I would like to use this library in vxworks and the linker in vxworks doesnt support PIC code. It complains about undefined _GLOBAL_OFFSET_TABLE_ symbol. I am wondering how I can obtain a non-PIC version of the library or if there is anyway i can manually set the _GLOBAL_OFFSET_TABLE_ to some value and fix the problem?

thank you,

loading the DIB into ipp

hi everybody
all these samples are beautiful but...
i need a simple starting point of:
1.take a direct show DIB.
2. find out what kind of ipp data it contains (channels , types , etc...).
3. loading it into an ipp buffer allocated and optimized with ipp.
because i am as green as can be with the ipp
samples will be appreciated in my mail

Image Subtraction

i captured the motion of player/ball from football movie by using image subtraction.
but the image quality is poor,
i used 3 images, one for source and another for destination and one more image for reference image..(see attachment)
anyone can tell me the good algorithm for image Subtraction...

Using IJL for MJPEG


I'm looking for the white paper titled "Using the IJL with JPEG Compressed FlashPix Files" as referenced in then ijlman.pdf for an example of extended baseline decoding.

I'm trying to implement an MJPEG decoder using the IJL and I can't figure out how to set up IJL for optimal decoding using persistent Huffman tables.

Are there any example code out there using IJL for MJPEG?

Help appreciated.

AddWeighted using a "third image" pixel weights


I'm looking for an IPP optimized mutation of the AddWeighted operation in which constant value weight would be replaced by the 0/1 pixel weights coming fromthe "third image" as follows:

image[x,y] = image3[x,y]*image1[x,y] + (1-image3[x][y])*image2[x][y]

assuming that "non-zero" value of compare operation below = 1???

image3[x,y] = ippiCompareEqualEps (image1[x,y], image2[x,y]);

Best regards,


Subscribe to Intel® Integrated Performance Primitives