Aliasing vs Noise

对于Aliasing和Noise之间的区别一直有所疑问，前段时间得到了一个解释：

Aliasing is a *specific* kind of error; it's when high frequencies show up as low frequencies.  It's like information from high frequency content "masquerades" as low frequency content, which is why aliasing is called aliasing.

If you think about the jagged edges on a line drawn on a computer, the discontinuity between the edge and the background is the highest frequency event possible -- an abrupt change like that has infinite frequency.  This shows up as the very regular, low frequency stair-step we all know and love.

Noise is unstructured, random error, and most anti-aliasing techniques are attempts to move the error from low-frequencies (which the eye is very sensitive to), to high-frequencies (which the eye can ignore under the right conditions).  Your eye is very good at averaging noise if it's high frequency enough, but low frequency phenomena stand out.

具体来说：Aliasing与Noise都是信号重建或是图像渲染时的一些error，两者之间的区别在于它们频率的不同，走样的频率较低，因而更容易被人眼识别，造成不适；而相对来产噪声的频率较高，一定程度上对视觉系统而言敏感度会比较低。

相关的应用：比如在渲染采样时的各种Sampling mode（比如stratified）就是一种通过合理的采样模式来将最终的走样转为噪声的方法；后处理渲染中在一些采样（如SSAO等）也会使用各种噪声来分散样点，减少走样进而提高最终的渲染效果。