Set English

Popis projektu

Môj dizertačný projekt, ktorý vedie Dr. Miloš Šrámek, pozostáva z troch hlavných častí:

• Návrh a implementácia novej reprezentácie voxelizovaných objektov kombináciou známych techník, ako sú limitované vzdialenostné polia a "run-length" kódovanie. Začlenenie novej objemovej reprezentácie do knižnice f3d.

• Modelovanie rozličných telies v tejto reprezentácii, so zameraním na tie, ktorých povrchy obsahujú ostré detaily (hrany a vrcholy v širšom význame). Vývoj nových pokročilých voxelizačných techník, ktoré berú do úvahy podmienky reprezentovateľnosti objektov, s cieľom vyhnúť sa tak renderovacím artefaktom, ktoré sú typické pre predošlé metódy.

• Využitie spomenutej reprezentácie pri simuláciách fyzikálnych javov, pri ktorých telesá menia svoj tvar v čase. Medzi takéto javy patrí napríklad topenie, sublimácia, zvetrávanie a podobne.

Veľká časť práce na prvých dvoch častiach projektu je hotová. Niektoré výsledky už boli publikované na medzinárodnej úrovni a ďalšie publikovanie je v príprave. Tretia časť projektu je vo fáze štúdia literatúry.

Predbežné výsledky

(táto časť stránky ešte nebola preložená do slovenčiny)

• A new volume represenation using modified version of run-length encoding has been proposed and implemented. The modification of the RL compression comes from the observation that the information about an object is stored just in voxels located in the surface vicinity if truncated distance fields are used. The typical row in the grid consists of several long sequences of outside or inside voxels separate with short sequences of transitional voxels. Therefor, our idea is to divide the row into segments of three types: outside, inside and transitional. For coding of an outside or inside segment we need to store just its length and flag (OUT, IN). In addition the transitional segment also stores a pointer at a list of voxels which form the segment. Whole the grid is stored as two dimensional array of compressed rows.

• The typical feature of discrete volume data is that the representation of fine details is limited by the Nyquist law. One goal of this project is to solve the problem of solids containing sharp details the characteristic dimension of which is under the Nyquist level and so they are not representable in the discrete volume representation. The main idea of our solution is to round these sharp details during the voxelization process in order to decrease surface curvature in the critical areas to a value corresponding with the Nyquist law. It means that objects are modified in a way to be representable in a grid with the given resolution.

  We have solved this problem for two categories of objects:

  • Solids obtained as a result of CSG operations

    
    
    More details in the paper CSG Operations with Voxelized Solids published in the Computer Graphics International Conference 2004.
    

    Abstract of the paper:

    We present a new technique for CSG operations with voxelized geometric objects, which are represented by truncated discrete distance fields supplemented by additional information about the surface normal. The technique removes artifacts of straightforward volumetric CSG operations by taking into account conditions for object representability, according to which sharp details are not correctly representable in discrete distance fields. The proposed technique solves this shortcoming by rounding edges and other sharp details. It works at the voxel level without the necessity for reconstruction of continuous object models.

  • Implicit solids with sharp details

    
    More details in the paper Representation of objects with sharp details in truncated distance fields published in the Volume Graphics 2005.
    

    Abstract of the paper:

    We present a new approach for voxelization of implicit solids which contain sharp details. If such objects are processed by common techniques, voxelization artifacts may appear, resulting, among others, in jaggy edges in rendered images. To cope with this problem we proposed a technique called Sharp Details Correction. The main idea is to modify objects during the process of voxelization according to the representability criterion. This means that sharp edges end vertices are rounded to a curvature, which depends on the grid resolution. Thus, we obtain artifact-free voxelized solids which produce alias-free images.