parallel image processing on cuda

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi All,

 I am having a Geforce GTX 460 and am planning to do some parallel image processing.

Problem def: I have an image A. I want to apply an image processing algo (for example lets say image dilation) for different regions of the image (the shape of the ROI is not fixed). How can i do this in parallel such that each ROI of the image is processed by different cuda blocks and hence speed up the process.

Thanks in advance for all the help.

Jacky,

2

Hi All,

 I am having a Geforce GTX 460 and am planning to do some parallel image processing.

Problem def: I have an image A. I want to apply an image processing algo (for example lets say image dilation) for different regions of the image (the shape of the ROI is not fixed). How can i do this in parallel such that each ROI of the image is processed by different cuda blocks and hence speed up the process.

Thanks in advance for all the help.

Jacky,

3

You could either execute the same kernel multiple times concurrently with different ROIs. Or, store all ROIs in some sort of array in global memory, and use blockIdx to select a ROI for each threadblock. This would be my best guess at this time.

4

You could either execute the same kernel multiple times concurrently with different ROIs. Or, store all ROIs in some sort of array in global memory, and use blockIdx to select a ROI for each threadblock. This would be my best guess at this time.