Changes

Here, we see that the program was improved by the optimizations of threads per block.

When the image side increases, the more efficient the kernel.

Each kernel individually found significant or marginal improvements after adjusting for thread/block size.