I tried the usual route using getImageData() and am getting between 50 ms for a small region and 100 ms for the full 2048 x 2048 canvas on a 8th generation mobile i5 in Chrome. In Firefox it is 50 ms and 150 ms, all on Windows 10. So the overhead alone of getting the image data from the canvas and writing it back can eat up most of the entire 50 ms budget if not exceed it. Or maybe I did it not in the proper way, I am totally not a web developer. There is also certainly some room for improvement in my implementation of the core algorithm. But unless there are more tricks than specifying willReadFrequently, it seem hard to get below 50 ms.
