The Content-aware Based Hierarchical Atlas Visualization

Hierarchical atlas data contains a large number of images visual information and hierarchical relationship among nodes, so it is difficult for the traditional atlas visualization schemes to express information directly and effectively. As we know, Image Collage technology can make full use of canvas to present the visual information of atlas, but it fails to express the hierarchical relationship between nodes. While Photomosaic technology is good at visualization effect but is poor in computing efficiency. This thesis focuses on offering an effective content-aware visualization scheme, which merges Image Collage technology and Photomosaic technology.The main research work and contributions of this thesis are as follows:1.Based on hierarchical atlas data, we designed a content-aware visualization scheme and built the level atlas visualization prototype system by integrating the Image Collage, Voronoi Treemap and Photomosaic. Image Collage and Voronoi Treemap are jointly used to present the hierarchical information and visual information, which can improve the efficiency of atlas browsing. For the target image chosen by users, an irregular Photomosaic visualization effect is provided, which supported a fine-grained global scalable atlas visual manner.2.By comprehensively applying color complexity, gradient complexity and image distinctness, we designed a content-aware image importance calculation method and an effective representative figure selection strategy for parent node. Since salient region plays an important role in the image, we computed the content-aware color complexity, gradient complexity and image distinctness as well as image difference degree. Meanwhile, we selected the image that has the smallest average difference as representative figure of parent node to reflect the characteristics of all the images in sub levels.3.According to the dynamic characteristics of hierarchical atlas data, we designed a two-stage stable layout scheme offering initial stable layout and node-added stable layout to enhance the availability of visual exploration. Initial stable layout algorithm uses hash functionand wachspress coordinate mapping to ensure the stability of the visual results at beginning stage.In order to reduce the influence generated by newly increased nodes, node-added stable layout algorithm applies fringe-insert strategy for new nodes, thus reduces the difference of adjacent visual status.4.Since typical Photomosaic technology with rectangular tiles has drawbacks as tedious visualization effect and obvious regular boundary, we offered an optimized irregular Photomosaic algorithm by rotation-relocation and fill-matching with acceptable computing time.This thesis extends general atlas data visualization to the hierarchical atlas data. Using content-aware color complexity, gradient complexity and image distinctness, we designed a reasonable calculation method of image importance and a selection strategy of parent node representative figure. Meanwhile, a stable layout scheme including initial stage and node-added stage has been designed to offer continuity visualization result. An optimized irregular Photomosaic algorithm is presented to enrich the visualization result. Finally, a new prototype system is developed.