Real-Time Color-Based Depth Cueing


This page contains some examples of hardware-accelerated color-based depth cueing. All images and movies were produced with our interactive rendering tool running on a Windows 2000 PC with with an AMD Athlon 650 and an NVidia GeForce3.


Technical Report

A detailed technical report on real-time depth cueing is available for download.


Video

The video demonstrates the use of various depth-cueing approaches within our OpenSceneGraph-based real-time 3D viewer. All sequences show interactive applications and were directly recorded to a camcorder via the S-VHS interface of the graphics board. A Windows 2000 PC with an AMD Athlon 650 and a GeForce 3 was used. Screen resolution was 800*600 pixels.

The first sequence shows non-photorealistic rendering of an engine block. The second sequence contains the cone tree visualization of hierarchical data. The video ends with a terrain rendering application. Hires still images of the scenes are shown below.

Regular Grid of Spheres

The following images demonstrate the effects depth-cueing for a simple test scene that consists of differently colored spheres on a regular grid. Image~(b) illustrates depth-cueing by fading to black.

Spheres Original test scene without depth-cueing. A hires image is available.

Spheres Test scene with intensity depth-cueing (black fog). A hires image is available.

Spheres Test scene with saturation depth-cueing. On one hand, the saturation depth cue is not as strong as the intensity depth cue (image above). On the other hand, a saturation gradient does not influence the intensity contrast and therefore allows to visualize even background objects. These far-away objects are not visible with intensity depth-cueing. A hires image is available.

Spheres Test scene with the combination of saturation-based and intensity-based depth-cueing. The dimming effect is only half of the desaturation effect, combining the advantages of both approaches: The depth cue is stronger than in mere saturation-cueing and, in contrast to intensity-cueing, background objects are still visible. A hires image is available.


Terrain Rendering

The following images demonstrate depth-cueing for terrain rendering. Here, depth-cueing is only applied to the terrain itself, the background sky and ocean keep their original colors.

Terrain Original scene without depth-cueing. A hires image is available.

Terrain Terrain with saturation depth-cueing. A hires image is available.

Terrain Terrain with intensity depth-cueing (black fog). A hires image is available.

Terrain Saturation depth-cueing is combined with a "sawtooth" function for intensity depth-cueing. The resulting image allows to identify lines of equal distance from the camera. In this way, depth-cueing even facilitates quantitative distance measures. This technique is related to classical contour lines (isolines) in topographical relief maps, but is more subtle. The "sawtooth" approach is especially useful for extended objects like terrains. However, it is less suited for scenes consisting of many small and separate objects because the context of lines of equal distance is destroyed by the blank regions between these objects. A hires image is available.


Non-Photorealistic Rendering of an Engine Block

The following images demonstrate depth-cueing for the non-photorealistic rendering of technical illustrations. Cool-to-warm tone shading is applied to enhance the recognition of surface orientation. A cool (blue) to warm (tan) transition of color tones indicates a change of the surface normal from left to right. Gooch et al.'s cool-to-warm tone shading is used [A. Gooch, B. Gooch, P. Shirley, E. Cohen, A Non-photorealistic Lighting Model for Automatic Technical Illustration, SIGGRAPH 1998 Conference Proceedings, pages 101-108].

This example demonstrates that saturation depth-cueing is close to Tufte's strategy of the smallest effective difference, making visual distinctions as subtle as possible, but still clear and effective. In particular, technical illustrations benefit from saturation depth-cueing because all structures - even in the background - are retained.

Engine Original rendering of the engine block without depth-cueing. A hires image is available.

Engine This rendering illustrates cool/warm shading and saturation depth-cueing for the engine block. The main differences to the original rendering appear in furthermost parts of the engine, especially in the upper portion of the images. Saturation depth-cueing causes rather subtle color changes. A hires image is available.

Engine This rendering illustrates cool/warm shading and intensity depth-cueing (i.e., black fog) for the engine block. The main differences to the original rendering appear in furthermost parts of the engine, especially in the upper portion of the images. Intensity depth-cueing harshly affects the image by completely fading away parts of the engine A hires image is available.


Information Visualization

The following image shows a typical application in information visualization. Here, a large amount of hierarchically structured data is given in the form of a mathematical tree and visualized by a so-called cone tree. The elements of the tree are represented by boxes and cones and connected by thin lines. These graphical objects are located and viewed in 3D space. Therefore, depth perception is quite important for the understanding of the tree structure.

Cone tree The original rendering on the left includes only the standard depth cues in the form of perspective and occlusion. With the saturation depth cue being enabled (right image), the depth structure becomes much clearer. The top and bottom images in the center of middle image show magnified details of the respective renderings. The in-between image provides a comparison between these two detail images, showing the differences of saturation levels of the two original images as gray-scale values. A hires image is available.


Last modified 11 May 2002
by Daniel Weiskopf