Ishita's profile - activity

signature of cpu ( c style ) cvResize is: cvresize(source array , destination array , interpolation method);

whereas ,signature of gpu resize is: gpu::resize(source array , destination array , destination image size , scale factor along x axis , scale factor along y axis , interploation method , asynchronous stream);

scale factors are the added ones.

why there is a difference in output of resize API among its cpu and gpu version ? why is the scaling factor argument added in the gpu version of resize?

I also observed the difference in cpu and gpu version of resize but still not able to figure out the reason behind it....

Does a cv::gpu API takes more time inside a for loop as compared to its corresponding C Style API in the for with same number of iterations.

i.e

` // GPU FOR LOOP

cv::gpu::GpuMat a , b; // assume they contain some data

for(int i = 0; i < 10180 ; i++)

cv::gpu::add(a ,b,a);

// CPU FOR LOOP

CvMat a1 , b1; //assume they contain some data

for(int i = 0; i < 10180 ; i++)

cvAdd(a1 ,b1,a1);

`

Timings:

GPU FOR LOOP : 0.87SEC

CPU FOR LOOP : 0.02SEC

why is it so ?