What's the difference between Aliasing and Noise?

[bugrunner]

Hi , All;

In Chapter 7 of PBRT, the stratified sampling is introduced to improve the quality of rendered image. The reason behind is that we can use carefully tuned sampling mode to change the aliasing into noise, cause human perception system is more sensitive to aliasing than noise. Then what's the difference between aliasing and noise? Anyone can explain these two concepts detailedly or give some references to check?  (not only in rendering but also in the signal processing)

Thanks!

[Greg Humphreys]

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.

--
You received this message because you are subscribed to the Google Groups "pbrt" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pbrt+uns...@googlegroups.com.
To post to this group, send email to pb...@googlegroups.com.
Visit this group at http://groups.google.com/group/pbrt.
For more options, visit https://groups.google.com/groups/opt_out.

--

Greg Humphreys

ɯoɔ˙lıɐɯƃ@ɹǝdɯnɥ

[bugrunner]

Thanks, Greg

在 2013年11月21日星期四UTC+8下午11时10分06秒，Greg Humphreys写道：

[tomwindcloud]

 

---

Hi，

    The second equation may be not correct on page 931(second edition).

   

    Bi(x) = sqrt(n),  (i - 1)/n <= x < i/n

 

Regards,

Xiong Wei  

[Matt Pharr]

Nice find--thanks! Updated errata will be posted online in a bit.

Regards,

Matt

--