Research On Pixel-based Mass Data Visualization

From the traditional data access technique to the present data visualization technique,data management realized a great step from the abstract to the intuitive. For the datasets,the data scale become increasingly larger, the storage structure become more complicated,and the data types also become no longer single. How to analyse data and then findknowledge and information visually from large numbers of data is particularly important.As the front direction of data visualization domain, pixel-oriented visualization techniquestands out from other visualization techniques with advantages of its effectively analysingand dealing with large scale database. This thesis carries on a research mainly from twoaspects of the time efficiency and display effect, aiming to find a pixel-orientedvisualization technique which has a simple structure, a convenient operation, a fastrealization, and an obvious display effect to the mass data visualization display.First of all, the thesis proposes a pixel visualization approach based on a scan lineseed filling technique. This approach combines two techniques of scan line seed fillingand pixel visualization, which uses the seed point to scan pixels horizontally in specifiedclosed area. Besides, the thesis chooses appropriate color model to realize mapping fromdata to color, and constructs a suitable pixel mapping model to realize pixel visualizationof large scale datasets, which has a higher efficiency and a better visualization effect.Secondly, the thesis proposes using quadtree technique to deal with the pixel array oflarge scale datasets. Use of quadtree technique to store and manage datasets which needsto consider the relationship between data nearby. It keeps the position attribute of data,and makes the process of the pixel mapping become easier.Thirdly, the thesis proposes realizing pixel visualization of the large scale datasetsdynamically, saving the pixel filling process as an avi movie. In this way, it’s convinientfor the users and data administrators analysing and observing the regulation of data change,and then tracking pixel filling process dynamically.Finally, according to the approaches above, techniques are evaluated on two realworld databases, Data Ⅰa nd Data Ⅱ.