Min Filter

Introduction

Given an image, we'll replace each pixel intensity with minimum value from its neighbouring pixels, where order of neighbourhood can be variable X ( > 0).

Performance

Prior to this writing, I wrote concurrent MinFilter implementation where each pixel was getting processed in different worker thread ( depending upon availability in thread pool ).
Now implementation has been improved by processing each row of image matrix concurrently.
Previously each thread was doing little work, but for large images ( & for higher order filtering ) communication cost ( work distribution ) was too high due to presence of large number of pixels in image.
So, I considered reducing number of works created, by processing each row concurrently on possibly different worker threads.
Check below for performance improvement, due to updating implementation