the 128x128 subsample method is noisy. i'm trying to resolve this by fixing possible issues using simulation.

i'm currently simulating upsample in javascript code.  ill post a link when web version is uploaded and article is up for the non float upsample method used. i thought it would have been uploaded a week or so ago but  it will be out this next week sometime. i hope to have it capable of 32x32 (the simulation) and eventually 128x128. this simulation code will help me fix noise issues at 128x128 and possibly greater upsample resolutions.

i have already included optical correction which compensates for individual sensor variation of angle. this effects and improves the pixel color accuracy. i included compensation on a hunch, and with trial and error it proved to be the correct thing to do. it slightly averages pixels together about  1/16 depending on specific settings in code. this is before subsampling is used. 

there is an other effect that i am going to implement that will make image upsampling have less shadowing and it is pixel offset calculation. depending on position of sub sampled pixel there should be a set amount of offset to its brightness value compared to its neighbors. 

i would have posted more results sooner, but i ran into this optical error, it can be viewed in both upsampled views and shows with a darker and lighter shadow in both 16x16 and 16x16 corrected. 

i have spend a little bit of time trying to see if i could make this better. granted in real life such a sharp contrast in backgrounds would most likely not be possible, but i spent time to see if there was something that could be easily done to make it look better. 

after bit of trial and error i found out that even with other upsample methods this specific sharp contrast and small size of image had issues with other upscale methods. so it is not just my method that has this issue, it still bugs me, but not as much as it did. part of the issue is no shadow information in image. this would not be an image in real life. 

here below is how it is output thru gimps upsampling algorithms from 8x8 grid to 16x16 grid, image was cut and pasted from screen snap shot of image scaled to screen 1600%

I will see how this algorithm works to see if improvements can be made in my subsample method.