Questions & explanations
1. What is global illumination in computer graphics?
Global illumination is a set of techniques that simulate how light travels in a scene, including both direct and indirect illumination. Direct light comes straight from light sources, while indirect light bounces off surfaces before reaching the eye. Global illumination accounts for color bleeding, where a red wall tints nearby white surfaces. Key concepts include radiance, which is the amount of light traveling in a direction, and BRDF (bidirectional reflectance distribution function), which describes how a surface reflects light. Realistic rendering requires global illumination to show subtle lighting effects. Without it, scenes look flat and unnatural.
2. Compare the advantages and disadvantages of using square markers vs. circular markers.
Square markers, like QR codes, are common because their corners make it easy to compute position and angle. They are simple to generate and detect. Circular markers are less common but can offer rotation invariance. However, circles are harder to estimate pose from because they don't have distinct corners. Square markers can be detected even when partly rotated in 3D, but their square shape may be confused with other square objects. Circular markers may be more robust to occlusion in some cases. Overall, square markers are more widely used due to simpler detection algorithms. The choice depends on the application needs.
3. What is Whitted ray tracing?
Whitted ray tracing is a recursive algorithm that simulates perfect reflections, refractions, and hard shadows. It starts by shooting a ray from the camera through each pixel. When the ray hits a surface, it spawns new rays: a reflected ray in the mirror direction, a refracted ray if the surface is transparent, and a shadow ray toward each light source. The algorithm continues recursively until a maximum depth is reached or the ray escapes. It produces crisp reflections and sharp shadows because it only handles perfectly smooth surfaces and point lights. Whitted ray tracing is the foundation of many modern ray tracers.
4. How is claymation different from traditional hand-drawn animation?
In hand-drawn animation, each frame is a separate drawing on paper or digitally. The animator draws every movement from scratch. In claymation, the animator works with physical clay figures that are moved and photographed frame by frame. Hand-drawn animation can be more fluid because drawings can be changed easily, but claymation gives a three-dimensional, textured look. Claymation takes longer because each tiny movement must be done by hand with the real model. Also, hand-drawn animation often uses many more frames per second for smooth motion. Both styles are artistic but require very different skills and tools.
5. Why is global illumination difficult to compute in real-time?
Global illumination is difficult because it requires tracing many light bounces across the scene. Each bounce adds more calculations, and the number of possible light paths grows quickly. Real-time rendering has to produce images in milliseconds, but full global illumination can take seconds or minutes per frame. Techniques like precomputed lighting or simplified models (such as ambient occlusion) are used to approximate global illumination. Even modern graphics cards struggle with complex indirect lighting at high frame rates. Therefore, games often use shortcuts and approximations to balance quality and speed.
6. How can we detect if a video is a deepfake?
Detection methods look for small mistakes that deepfakes often make, such as inconsistent blinking or unnatural eye movements. Another clue is poor blending at the hairline or changes in lighting across the face. Some detectors analyze facial movements that are different from real human patterns. Machine learning models can also be trained to spot deepfakes by examining pixel-level artifacts (tiny errors left by the generation process). Additionally, checking the video's source and using blockchain verification can help confirm authenticity. No method is perfect, but researchers keep improving detection tools.
7. How does connecting a camera path and a light path reduce noise?
By connecting a camera path vertex to a light path vertex, you create a complete path that includes both the light's history and the camera's history. This gives the algorithm more chances to find bright contributions, especially for paths that would be rare in standard path tracing. For example, in a scene lit by a small window, the light path may find the window, and the camera path may find the spot being lit, and connecting them directly yields a bright sample. This reduces variance because many paths that would otherwise be dim are now brighter. The result is less noise for the same number of samples.
8. Why might HDR images sometimes look unnatural?
HDR images can look unnatural when tone-mapping is applied too aggressively. Tone-mapping compresses the wide brightness range to fit a standard screen, which may flatten contrast or create halos around edges. Over-enhancement can make shadows too bright and colors too saturated. Ghosting artifacts occur if objects move between exposures, causing blurry or duplicate edges. Some HDR software boosts local contrast excessively, giving a gritty or cartoon-like look. The goal is to balance detail with natural appearance. Subtle HDR processing yields realistic results, while heavy processing looks artificial.
9. Give an example of a scene where bidirectional path tracing is much better than standard path tracing.
A classic example is a room lit only by a small light source outside, with a narrow window. Standard path tracing requires the camera path to randomly hit that tiny window, which is very unlikely, causing dark noise. Bidirectional path tracing finds the window from the light side and connects to the camera side, making it easier to get bright samples. Another example is a glossy surface reflecting a bright light; bidirectional path tracing can directly connect the light path to the glossy vertex. These situations cause severe noise in standard path tracing but are handled well by bidirectional methods.
10. How does lighting affect silhouette animation?
Lighting is the most important part of silhouette animation. A bright light shines from behind the screen, making the cutouts dark and the background light. The position and strength of the light change the look of the shadows. If the light is placed directly behind, the silhouettes are sharp and clear. Moving the light to the side can create softer edges or partial shadows. Animators can also use colored lights to create different moods. The light must be steady so that the brightness does not change between frames. Without careful lighting, the silhouettes might not be visible or the effect is lost.
11. What is a challenge in implementing foveated rendering?
A major challenge is accurate eye tracking. The system must know exactly where the user is looking to set the high-res area. If the eye tracking is slow or incorrect, the high-res area may shift away from where the eye actually points. This causes a noticeable blur in the center of vision. Another challenge is the transition between high and low resolution regions. If the change is abrupt, users see a distinct ring. Smooth transitions require careful blending. Also, foveated rendering adds latency because eye tracking data must be processed before rendering. Overcoming these issues is active research.
12. Compare optical stabilization with digital stabilization.
Optical stabilization uses moving lens elements or a shifting sensor to physically counteract shake. It works in real time without cropping the image, providing maximum field of view. Digital stabilization processes the video after capture, cropping and warping to smooth motion. Optical is generally better for high-quality footage because it doesn't reduce resolution. Digital stabilization can be applied to any video but may introduce artifacts and loss of detail. Many cameras combine both: optical first, then digital for extra smoothing. Digital stabilization is cheaper and can be added in software.