Dilation & Erosion for c6400

The input data is in bit packing format, which means each bit in the first byte represents a pixel.

That means that "colsper8" is "capture_imageCols" divided by 8, isn't it?

The thing is, the function seems to be doing nothing. If I declared another buffer "pingPongBufferC" and initialise it with zeros, I would get a totaly blank/black screen with this call:
 IMG_dilate_bin(pingPongBufferB, pingPongBufferC, deMask, colsper8);

Test program is provided with the library, I think you should run the test program successfully before you
begin to use the library function.

I did. I always do. If you know what's the problem please say so. It may be obvious to you, but not to me.

Your description of it not doing anything sounds like a cache coherence issue.  Is your buffer in external memory?  Do both CPU and EDMA access the data, i.e. CPU for the processing and EDMA to output the image?  If so, after you complete your processing you should do a block writeback of that buffer so that the updated data will land in the SDRAM for the EDMA to read.

I don't have external memory, everything is in SDRAM. As for the 2nd question, yes, both the CPU & EDMA access the data. I have a DAT_copy function (to copy to display_FrameBuffer) after all processing is done. So the order is like this:

1. threshold
2. dilation
3. copy to display

Another weird thing is if I replace the dilation in step 2 with Sobel (also from IMGLIB), the program works as it should. Therefore I do not think it is a cache coherence issue (sorry I forgot to mention this in the first post).

No one has other ideas?
Hanief