| Summary: | 碩士 === 國立政治大學 === 資訊科學系 === 107 === Lego® has been developed and well-known since 1940s. In computer science, it has been studies that describing the procedure to automatically generate Lego sculpture, including 2D and 3D. However, these studies mostly only consider basic Lego brick, which has cuboid appearance, as constructing component. As a result, the generated Lego sculpture appears pixelized or voxelized. We propose a method to improves the smoothness of the contour of 2D Lego sculpture by adding smooth Lego parts, including different shape and size. Due to the expansion of constructing component, the considering of similarity between the Lego sculpture and input image during the constructing becomes necessary. We introduce the concept of using potential field to determine the similarity between Lego and image, by calculating the distance and the variation of contour between Lego and image. Variable shape and size of Lego parts leads the possible combination of Lego sculpture growing exponentially. It is impossible to find an optimal solution of combination in polynomial time by brutal force. We use an iterative strategy to generate an approximate solution. Choosing the brick that maximally increases the similarity in every round of iteration. For internal area of contour, we using basic brick to increase the stability of Lego structure. Finally, we using variable image as input to examine the efficiency, stability and scalability of our method.
|
|---|
