going to 64x64 sub sampling i had to fix some artifacts. this was done with some tricker in the image processing, but i want to see more about what is or is not going on. i'm creating a java program that will plot out differences in images from different sub sample methods. the goal is to improve color accuracy of pixels when iterated several times. it is important for 128x128 since this uses no buffer and all data in the end is generated from upscales of original data and side pixel information. the non floats method i use is less blurry but color differences per 8x8 or 16x16 space cause blockiness. i'm trying to resolve why this happens, and also have data for an article soon to be uploaded that shows how to quickly upsample data with almost no loss in performance, or memory footprint. it is more impressive to those that actually will use it. it has very little memory overhead and it is several times faster than other versions of sub sampling.