A problem with ippiMean_8u_C1R

[amir_l_]

Hello,

I have a library written in C compiled with IPP in Linux (Ububtu 12).

In the library I calculate the mean value of an image. There is a specific image that it's average is NAN. This function is part of a bigger loop that processes images. The input image's memory area is fixed and doesn't vary from image to image.

I verifired that all the image values are valid, the ROI is valid and the sizes are valid: all the input parameters are valid.

The return code is success and the mean value itself is NAN.

If I calculate the mean value with my own code the result is right and valid.

Please advise

Thanks

[SergeyKostrov]

>>... >>The return code is success and the mean value itself is NAN. >> >>If I calculate the mean value with my own code the result is right and valid. >>... I did not understand where the problem is. You've stated that everything is valid and return codes are success for both test cases. Could you specify what kind of problem you have with ippiMean_8u_C1R function?

[Gennady_F_Intel]

or please give us the test which we can compile and check the results on our side.

[amir_l_]

The code works well 99.9% of the time.

In rare cases the function returns a success code but the mean value returned is NAN.

This code is part of a library. I have two modes of running my scenario:

1. A simple test application that tests the library

2. A bigger project that the library is staticaly linked to

I ran the same scaenrio in both ways (and verified that the inputs to the mean function are exact) and the bigger project's path have the reported problem. The only difference between the two calls is the image's address. In both cases it is a global array.

It's not a simple peice of code that may be tested easily on your side.

[Chuck_De_Sylva]

Amir,

Also, what are the input values you are using?

- Chuck

[amir_l_]

srcStep=112, ROI={4,4} and the image has a constant pixel value of 17.