地理資訊系統中的階層式預看影像及無失真影像壓縮法

Abstract

在地理資訊系統中,有大量的影像其尺寸超過電腦螢幕的解析度。現在一 般的電腦軟體都是使用捲動軸來捲動螢幕,但是這種方式通常是相當沒有 效率的。本論文提出了一個階層式影像顯示系統,它能將一張影像中使用 者有興趣的部份很有效率的抽取並顯示出來。這個系統以兩大步驟來達成 這個目的。第一:從原始影像中抽取許多尺寸不同的子影像。第二:利用這 些子影像來完成階層式預看影像的功能。本系統顯示影像的程序如下:先 顯示一張尺寸為最小的子影像,然後由使用者決定是否對這張子影像有興 趣,若答案為非,本系統即載入另一張子影像或是停止本系統。若答案為 是, 使用者決定想看該子影像中的那一個部份, 本系統即將該部分以較細 緻的影解析度顯示出來。本系統一直重複上述的動作, 直到使用者己經獲 得他想要的資訊為止。本論文的第二個重點是對地理資訊系統中的影像給 予適當的壓縮。對於二元影像本論文提出一種經過修改的群長壓縮法來壓 縮,這個方法首先求出以多少位元來表示該影像中所有的群長是最有效率 的,然後即利用該資訊來壓縮該影像。對於地理資訊系統中的灰階影像,本 論文使用無失真預測編碼法及最佳變動長度編碼法來壓縮。無失真預測編 碼法將每一個像點的值用它鄰近像點值的線性組合來做預測。藉此它可以 將原始影像轉成另一張預測錯誤影像。然後使用最佳變動長度編碼法來對 預測錯誤影像做壓縮, 可以達到不錯的壓縮比。 There are many large images, with their size exceeding resolutions of a computer monitor, in the geographical information system (GIS). Available application softwares usually design scroll bars to scroll the images. But this method is generally not quick enough. In the thesis we present a hierarchical image display sytem which can display a region in an image that a user is interested in very quickly. This system contains two operations: extracting previewed subimages from an input source image, and displaying the images hierarchi- cally.When a use intends to view an image, we show the lowest level previewed subimage. If user decides to check more details of a certain region then the corresponding region in a higher level will be shown. This procedure continues until the shown region is satisfied or the user terminates the viewing operation. To store geographical images efficiently by properly compressing them is the second concern of this thesis. For bilevel images, a revised run length coding scheme is presented. It calculates the number of bits required to encode each run in an image at first. Then it uses this number to encode each run length in the input image. For gray level images, a combinational scheme of lossless predictive coding and optimal vairiable-length coding is used. The lossless predictive coding predicts the intensity of a pixel by linearly combining the intensities of its neighboring pixels. Thus an input image can be transformed into a prediction error image. We then apply the optimal variable-length coding scheme. A good compression ratio can be obtained.