Made voluminal direct

The direct voluminal Rendu is a technique used to post a projection 2D of a set of data 3D.

Made Voluminal Direct

Made voluminal direct requires that each value sampled within volume is associated with an opacity and a color.

That is carried out thanks to what one calls a transfer function transfer F

F (sample_value) - > (color, opacity)

That Ci is generally modelled by a linear function by part (two-dimensional board containing the values of the angular points of the function) but also by all other types functions (function in staircase,…).

The values of opacity and color are converted into components RGBA (channels red, green, blue for the color, and channel alpha for opacity) and the resulting compound is projected on the corresponding pixel of the frame buffer. The way in which that is fact depends on the technique of returned. These techniques can be combined in order to benefit from the various advantages which they comprise.

For example an implementation of the type shear sharp can use the routines of texturage material of the graphics card in order to draw the aligned sections.

The Throw of rays

See also: To launch ray

The most common means to project an image 3D is the throw of rays through volume. This technique consists in generating a ray for each pixel of the desired image.

Generally the ray is sampled with regular step inside volume and the values of colors RGBA of the pixels thus calculated are summoned in the frame buffer in proportion to their contribution of Alpha transparency.

This type of algorithm is easily parallélisable on architectures SIMD.

Splatting

This technique compromises the quality of returned throw of ray to gain there in speed of treatment: Each under voluminal element of volume to be returned is projected directly in the screen plan by using a representation 2D appropriate to these under elements (generally ellipsoidal) whose properties of color of transparency gently vary center towards the outside of projection.

According to the application, representations 2D with the uniform properties are also used.

Shear Warp

It is an new approach of made voluminal developed by Philippe Lacroute and Marc Levoy. In this technique, volume is transformed so that the emitted rays are aligned on the axis of the normal of the face of volume nearest from the point of view. Each section of volume thus undergoes a translation and a scaling.

Thus the sections of volume (textures 2D) are accumulated successively moreover nearer with most distant by respecting the contribution from each sampling to the final pixel. This intermediate image is then transformed to respect the deformations related to the prospect from the point of view.

This technique is relatively fast at the price of a less precise sampling and thus of a returned potential of bad quality compared to the traditional throw of rays.

It is used in order to:

  • to reduce the cost:

  1. of projection of the algorithms of returned directed objects.

  2. of the throws of ray of the directed algorithms image.

Plating of texture

The plating of texture on geometrical objects is today carried out in a material way by the majority of the three-dimensional chart systems. It is thus easy to visualize sections of volume for a moderate computing time.

These sections can be aligned with volume and returned with an orientation depending on the position from the point of view. This type of visualization brings a great speed but to the detriment of one returned of average quality.

Another type of visualization consists in D-sampling volume by sections aligned with the point of view. One gains thus in quality of returned but with the detriment of the cost of D-sampling of volume to each image.

Returned Accelerated by the graphic material

A technique recently exploited to accelerate returned is the use of modern graphics cards to accelerate the traditional algorithms like the throw of rays.

The appearance of the shaders sounded the advent of the GPU like a very cheap and ideal architecture SIMD to carry out all kinds of parallel calculations.

The pixels shaders are able to read and write very quickly the memory of texture, to carry out logical and mathematical operations basic.

They appear of a great utility during heavy treatments such as the throw of ray and more generally the treatment of the signal.

With OpenGL version 2.0, the shaders are now able to function as of architectures MIMD (they are capable of independent connections) with 48 parallel treatment units using until 1Go of memory of texture and of the numerical formats of a high degree of accuracy.

Such a power increases in a drastic way the speed of execution of algorithms like the throw of ray, making it possible even to use them for returned in real-time!

See too

Random links:Treaty of Nankin | Jacques Dutronc (1970) | Walter Fürst | Park of Poissy - Large-Girdles | Auctoritas principis | Médaille_distinguée_de_vol