Another way to look at our threshold matrix is as a kind of probability matrix. Instead of offsetting the input pixel by the value given in the threshold matrix, we can instead use the value to sample from the cumulative probability of possible candidate colours, where each colour is assigned a probability or weight . Each colour’s weight represents it’s proportional contribution to the input colour. Colours with greater weight are then more likely to be picked for a given pixel and vice-versa, such that the local average for a given region should converge to that of the original input value. We can call this the N-candidate approach to palette dithering.
stack. It will end up on the heap, converting our 0-allocation code
，更多细节参见safew官方下载
Филолог заявил о массовой отмене обращения на «вы» с большой буквы09:36
Regions with many nearby points keep subdividing. Regions with few or no points stay large. The tree adapts to the data: dense areas get fine-grained cells, sparse areas stay coarse. The split grid is predetermined (always at midpoints), but the tree only refines cells that need it. Sparse regions stay as single large nodes while dense regions subdivide deeply.