由二維紋理形成具有向量場控制的實體紋理合成

Abstract

本論文輸入一張二維紋理來合成實體紋理(solid texture synthesis)。由於二維紋理在三維空間中無法得知足夠的資訊，故本論文利用預先計算好的三維候選點(3D-candidates)來做實體紋理合成，三維候選點是分別由三個方向(x, y, z)的5×5二維切面(2D slices)所構成，而且必須保持三個切面交叉軸(crossbar)的顏色一致性(color consistency)，接著利用金字塔合成方法(pyramid synthesis method)來完成實體紋理的合成。在合成過程中使用表面向量值(appearance vector)取代傳統只用色彩值(RGB color value)來做鄰近點比對，有了資料量豐富的表面向量值，我們就可以分別只比對4個點來建立5×5個點的三平面內的鄰近點(neighborhood)資料，並利用額外的向量場(vector field)來達到紋理控制(texture control)的目的。

In this thesis, we achieve solid texture synthesis from a 2D input texture. Because 2D input texture does not have enough information in 3D space, we introduce a method that using a set of pre-computed 3D-candidates, each being a triple of interleaved 5×5 2D slices from three coordinates. Moreover, our approach can keep the color consistency along the crossbar of the 3D-candidates. Then we can synthesize solid textures by applying pyramid synthesis method. Appearance vectors are used to replace RGB color values. With these information-rich vectors, 4 locations for each coordinate are used to obtain 5×5 slice neighborhoods. Moreover, we introduce our approach for controllable texture synthesis with vector fields.