Hello
I want to partition a wide image (e.g. 4000x3000) into a number of blocks and then perform my algorithm on each block in parallel. Since I may lose some information in the border of blocks, I need to do it with an overlap percentage. Regarding this, I have 2 questions:
1- Is there any c++ implementation considering that overlap percentage?
2- In terms of performance and cache, is it better to partition the image into rectangular wide blocks, or square-shaped blocks?