RE: Linear Approximation - Introduction and Examples
You are viewing a single comment's thread:
Cool, could you do a video on something I had to think about in order to reduce colors in an image, look at the following saying:
- There is 20K colors you can know for-each their usage based on the sum of possible usage weight.
- A color is a red value (0-255 included), same for the green, blue, and alpha (opacity value), in radix 2 it makes them a length of 32 (bits), which also is named 4 bytes.
- To avoid comparing 20K colors with themselves (20K*20K) is loud, we sort them in cluster, sometimes of 4, 8 12, bits which gives us a number of cluster that is ranging from 1 to 4096 because we reduce the colors like a gravy we encode them on 16 bits (which gives us 4096 clusters, 0-4095 included) and that makes us way less operations
Then my question is: given those concepts (my code here, https://dev.to/vipert/quantimat-reduce-of-20x-some-colors-within-88ms-down-to-2000-2pho) what are called such a reduction passing through binary values in mathematics?
0
0
0.000
Interesting. I had not worked on such a problem before. But searching online, I believe this "K-Means method" may be what you are looking for: https://muthu.co/reduce-the-number-of-colors-of-an-image-using-k-means-clustering/
Yeah it is similar it works with defining kind of clusters too yet it happens to add random variables and add similar entry around those random points, rather with binary reduction it's fast and deterministic but you have a lot of blending operation to do per clusters...