Abstract
The main research topic of the thesis is on the analysis, visualization, and synthesis of artists’ 2D image editing and 3D modeling operation sequences. The thesis shows that the additional information of editing history can improve traditional computer graphics algorithms and enhance the usability of UI of graphics design software. The thesis describes three major components: a nonlinear revision control system for 2D image editing software, a data-driven adaptive history list and a history-assisted view- point selection algorithm. The nonlinear revision control system transforms the linear edit- ing histories recorded on GIMP into a nonlinear DAG. Based on the DAG, user interfaces and algorithms for revision add, revision branch, revision diff and revision merge for im- ages are proposed. By further analyzing the editing histories, the thesis describes a data- driven adaptive history list that improves the traditional editing histories user interface with the function of semantic navigation and smart undo. Former is based on a SVM classifier while later a n-gram algorithm. Finally, the thesis describes a history-assisted viewpoint selection algorithm. The algorithm segments the 3D model into different regions based on time map then select the best viewpoints for segments by analyzing the camera history.