| Summary: | 碩士 === 國立暨南國際大學 === 資訊工程學系 === 97 === The 3D graphical representation of models on computer is becoming more and more accurate today. These representations often require complex calculations, but because computer hardware is now in a transitional stage, the overall efficiency may be affected when graphic calculations overloading the hardware on the hardware. The use of Level of Detail (LOD) can improve efficiency in graphic systems; the level of simplification can be adjusted according to system requirements, allowing graphic rendering to function more smoothly. However, as the complexity of models is reduced, it becomes more difficult to identify them. How to preserve the characteristic features of model is an important problem.
To aid in identifying models, research has begun examining a visually based LOD. The goal of this approach is to simplify models in accordance with human perceptual judgments. This thesis attempts to propose a simplification method based on human perceptual psychology. We start with the idea that everyone will have the perception of the models. This perceptions we call Abstraction, we use the faces of Abstraction as a basis for restricting the limits of model simplification.
This thesis proposes a cognitive oriented LOD system based on Abstraction. An Abstraction is a model composed of simple geometric shapes, which represents the characteristics of the original model. This simplified description allows a model to retain its characteristic feature when it is simplified. We then combine the algorithm used to simplify models (QEM) with a weight set by the Geodesic Distance sum to Protrusion values to preserve the relatively protrusive surfaces. In short, we optimize faces of a model by simplified description of Abstraction.
|
|---|
