Lecture 9 Flashcards
What is visual coherence?
Visual coherence is when our virtual model doesn’t look out of place in the real world.
How can we allow our virtual objects to be occluded with phantom rendering?
We can perform phantom rendering, to do this, we create a virtual model of something in our real scene and render it invisibly, we can then not render anypart of another model that is behind our phantom model. The steps for this are:
1. Draw video.
2.Disable writing to color
buffer.
3. Render virtual representation of real scene.
4. Enable writing to color buffer.
5. Draw virtual objects.
What is probabilistic occlusion handling? Why is it useful?
Default phantom rendering requires an accurate model, tracking data and registration, this can be difficult. Probabilistic occlusion handling creates a soft transition between the occluding and hidden object to compensate for tracking and registration error, the occluder will be more transparent if the occlusion probability is lower.
What can sparse models be good for?
We can create sparse models from geographic information systems (like google maps) to allow for segmentation of the image to be used as depth cues.
What does an online depth map do?
Uses depth map from a stereo camera or RGBD to compute correct occlusions.
What are some things we could do to simulate camera artifacts?
Add radial distortion by rendering to a grid, half sample and mix colours, add gaussian blur, add a motion blur based on tracking. Do Bayer sampling to seperate the colour channels and add noise, add a horizontal blur, then sharpen with a sharpening mask, then blend, recombine and convert to RGB.
How can we do realistic lighting with image based lighting? What is another possibility.
Capture an environment map using a reflective sphere within the scene, this can be used to create reflections, the light sources can then be extracted from the environment map to allow for shadow rendering.
Another possibility could be to do light estimation using an RGB-D camera.
Why might we want X-Ray visualisation? How can this be done?
We may want to render something that is behind another object without making it look like it is in front. This can be done by extracting important image elements like edges, salient regions, or high textured regions and using them for occlusion.
What are superpixels?
Superpixels are merging of pixels into superregions of similar pixels.
What is an x-ray cutaway?
Cut into the occluder to reveal fully opaque hidden objects.
What are visualisation metaphors?
Symbols used to indicate something within an image.
