A Tool for Adaptive Lighting Design

This research introduces an adaptive lighting design tool for 3D interactive environments, designed to solve the challenge of maintaining aesthetic lighting quality despite unpredictable user interaction. While traditional media like film and theater rely on static, manually adjusted lighting, interactive games require dynamic systems that can respond to runtime changes. The proposed system builds upon the Expressive Lighting Engine (ELE) by incorporating image-based analysis and perceptual lighting theories to better align with designer intentions and object material properties.

The methodology centers on a design-by-example approach, where artists provide an image representing their desired aesthetic. The system analyzes this image to extract lighting direction and intensity contrast metrics, utilizing psychophysical models such as Single-Scale Retinex for color constancy. To facilitate real-time performance, the system pre-calculates a five-dimensional response surface for objects, which maps camera direction, key light direction, and fill intensity to specific shading and contrast outcomes. At runtime, the system uses stochastic gradient descent to adjust lighting parameters, balancing natural scene illumination with the target aesthetic defined by the example image.

This work addresses significant limitations in previous automated lighting systems, which often relied on unintuitive numerical constraints or failed to account for the visual impact of object materials. By shifting toward a more intuitive, image-based interface and integrating perceptual models, the system aims to provide a more robust and artist-friendly solution for game developers. The scope of the research is focused on real-time 3D rendering, specifically optimizing lighting for individual objects or sets of objects within a scene to ensure visual continuity and artistic integrity during gameplay.