<%@ include file="../../check-login.jsp" %> Subdivision.org: Re: Loop Adaptive Subdivision Problem <% session.setAttribute ( "messageNum", new Integer ( 1625 ) ); %>

Re: Loop Adaptive Subdivision Problem

Posted by Scott Schaefer on March 02, 2005 at 16:56:42:

In Reply to: Loop Adaptive Subdivision Problem posted by suneel on March 01, 2005 at 13:01:08:

: The algo works by finding an error and divides the edge and neighbours by loop. But the thing is when i do it on a cube, it dont get a sphere, like the traditional loop scheme. Instad i get a pebble with blisters.

First off. You mean a sphere-like object, right? Loop subdivision doesn't reproduct trig functions and cannot generate an exact sphere. Usually you get some smooth round shape in the end. A pebble with blisters sounds undesirable though.

: At this point i want to know if adaptive subdivion will ever give me a sphere? cause it seem like if your adaptively subdividing a mesh, you disturb the symmetry and loose the shape.

Well, I'm not exactly sure what your algorithm is or how you split shapes with it to generate the multi-resolution framework. However, you should notice that you cannot just apply the rules hap-hazardly and expect to get out the same shape or even something similar. Subdivision is a parametric method meaning that it doesn't know anything about the space that you embed the object in. Therefore, a geometric criteria may be difficult to use to generate an adaptive subdivision scheme. Generating any type of adaptive scheme that has guaranteed bounds from the true surface may be quite difficult.

However, that's what I would like to see out of an adaptive scheme. It's okay if you don't generate the exact same surface as the original. However, I would want bounds saying that the adaptive surface is no more than X away from the true surface. This type of property could be used to measure error in projected screen space for instance.

: So does it mean adaptive is good for meshes that already have a defined shape towards their limit surface?

Hmm... I don't think I understand this question.

Follow Ups:

Post a Followup

Name:
E-Mail:

Subject:

Comments:
: : The algo works by finding an error and divides the edge and neighbours by loop. But the thing is when i do it on a cube, it dont get a sphere, like the traditional loop scheme. Instad i get a pebble with blisters. : First off. You mean a sphere-like object, right? Loop subdivision doesn't reproduct trig functions and cannot generate an exact sphere. Usually you get some smooth round shape in the end. A pebble with blisters sounds undesirable though. : : : At this point i want to know if adaptive subdivion will ever give me a sphere? cause it seem like if your adaptively subdividing a mesh, you disturb the symmetry and loose the shape. : Well, I'm not exactly sure what your algorithm is or how you split shapes with it to generate the multi-resolution framework. However, you should notice that you cannot just apply the rules hap-hazardly and expect to get out the same shape or even something similar. Subdivision is a parametric method meaning that it doesn't know anything about the space that you embed the object in. Therefore, a geometric criteria may be difficult to use to generate an adaptive subdivision scheme. Generating any type of adaptive scheme that has guaranteed bounds from the true surface may be quite difficult. : However, that's what I would like to see out of an adaptive scheme. It's okay if you don't generate the exact same surface as the original. However, I would want bounds saying that the adaptive surface is no more than X away from the true surface. This type of property could be used to measure error in projected screen space for instance. : : So does it mean adaptive is good for meshes that already have a defined shape towards their limit surface? : Hmm... I don't think I understand this question.

Optional Link URL:
Link Title:
Optional Image URL: