Transparency and translucency

Abstract: We present a new method based on GPU acceleration for real-time transparency and translucency rendering. Our method computes refraction at both the front and back sides of a transparent object, as well as internal reflection, thus delivering interactive realistic transparency effects on a commodity PC. The real-time performance is made possible by a new acceleration data structure, called geocube, that enables the use of GPU for fast ray-surface intersection testing. In addition, within the same framework, we introduce the novel use of the mip-map for a hierarchical representation of a sequence of key prefiltered environment maps to simulate translucency. By taking ray depth into account and using GPU to interpolate the key filtered maps to produce the desired blurring effects, we achieve real-time realistic translucency rendering of slightly scattering media that allows show-through of background details.

Abstract: PROBLEM TO BE SOLVED: To provide transparent or translucent matter to be printed which enables a backside design to be lightly seen deep in a surface design when seen from the surface side, and also enable a surface side design to be lightly seen deep in the backside design when seen from the backside, without deteriorating the transparency and translucency. SOLUTION: A surface design 4-1 is directly printed on a transparent or translucent shrink label 2. A concealing layers 4-2, 4-3 combined with a light transmission and a light reflection functions are printed on the surface design 4-1. Then, a backside design 4-5 is printed on the concealing layers 4-2, 4-3 on the assumption of being seen from the backside of the label 2. This allows the concealing layers 4-2, 4-3 to control the light transmission quantity and the light reflection quantity, making the most of the transparency and translucency of the label 2. Thereby, the backside design 4-5 can be lightly seen deep in the surface design 4-1 when seen from the surface side, while the surface design 4-1 can be lightly seen deep in the backside design 4-5 when seen from the backside. COPYRIGHT: (C)2006,JPO&NCIPI

Abstract: When a surface is viewed through a partially-transmissive material, the optical contributions of the two layers in a given viewing direction are collapsed onto a single intensity in the projected image. If a computer vision system is to recover the scene correctly, it must be able to decompose or scission the image intensity into the separate contributions of the two material layers (see Fig. 1A). The inverse problem of recovering layered surface structure is generally divided into two sub-problems: (1) the qualitative problem of inferring the presence of an interposed partially-transmissive layer in parts of the image; and (2) the quantitative problem of assigning surface properties—reflectance and transmittance—to the separate layers.