Knowing this, we can modify the N-Convex algorithm covered earlier such that the candidate weights are given by the barycentric coordinates of the input pixel after being projected onto a triangle whose vertices are given by three surrounding colours, abandoning the IDW method altogether1. This results in a fast and exact minimisation of , with the final dither being closer in quality to that of Knoll’s Algorithm.
别管什么开笔要文采飞扬，别想什么成篇要结构精妙，就一个字：写。像和老朋友聊天那样放松地写，像在酒桌上给人讲故事那样洒脱地写。让文字先流出来，哪怕是杂乱的水，也得先有个池子装。
。关于这个话题，51吃瓜提供了深入分析
+max_retries: int
mark candidate[n] as used
圖像來源，Getty Images