Systemic Single-frame Comics Vectorization and Screentone Manipulation and Real-time Rending

• Main Authors: Guo-Wei Lee, 李國維
• Other Authors: Chih-Yuan Yao
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/57038062899066468329

碩士 === 國立臺灣科技大學 === 資訊工程系 === 103 === Manga are a popular artistic form around the world and artists use simple line drawing and screentone to create all kinds of interesting productions. It is important for digital reproduction to reproduce these elements for proper delivery of contents and intentionsㄛ. Additionally, popular Manga are usually reproduced as Cartoon animations and it is tedious to animate raster images. Vectorization can be helpful for both Manga reproduction and Cartoon production. However, scanning methods still dominate Manga digital reproduction and the results are not satisfied because of noises and limited resolution in addition to a large amount of man power. Therefore, this proposal aims at transforming a scanned Manga object to a vector representation for interactive manipulation and resolution independent rendering. The process can be progressed in the following steps: 1). Scanned Manga pages are transformed to its corresponding black-and-white format. Manga pages are generally scanned in gray format to maintain the screentone details. Although binarization can make Manga features distinct, it is not a trivial job to determine a global or local threshold to properly maintain screentone details. Therefore, we aim at designing a binarization method for Manga. 2). Manga elements are detected and identified along with their corresponding types and properties. Generally, elements can be categorized as strokes and screentone. However, screentone patterns are chosen based on authors’ preference, mood and the requirement of scenarios and therefore, it varies too much to be really difficult in robust detection and identification. We plan separate the process into two sub-steps: detection and identification. A algorithm is designed to locally determine the screentone regions and then another method extract the pattern in the region to look up in our collected screentone database. 3). These detected and identified Manga elements must be vectorized for reading with different scales on different devices of different resolutions and screen sizes. For strokes, a ii clear mechanics must be designed to remove mixing-up among strokes and screentone regions for smooth and distinct strokes. Then, the strokes can be represented as fat spline lines. Furthermore, screentone regions can be triangulated along with its screentone type and properties for later procedural rendering. 4). Finally, both represenations must be rendered according to the reading conditions such as screen size, screen resolution, and reading scale. Clearly rendering screentones is even more important to express the intention of authors. We plan to design different procedural pixel shaders for different types of screentone patterns and a curve-based shader for strokes. Additionally, after vectorization, we also plan to design several interesting manipulation tools such as modify the screentone pattern and density and deform the vectorized characters and objects for interesting effects to ease the generation of Manga and Cartoon.