影像文件處理

 

影像文件處理

Image documentation

簡  介

 As we know, the problems of image documentation have existed in our lives for while. No matter in scanning techniques, in printing skills, or in the processing of documentation, image documentation has been all around us. For these reasons, we have concentrated on this subject for improving the performances of image documentation. Thus, I simply listed a few main topics that our group will work on in the future as below:

1. Pre-processing of document images:

 Before performing image documentation, we always have to pre-process the original image for future easier uses because of some noise in the image. In this part, there are several valuable but difficult problems for us to straighten out such as de-shading, auto-cropping, de-bleeding, etc.

2. Segmentation of Document Image:

 We can further go on the next steps of image processing more easily if we have smarter pre-processing of image documentation. That is why the level of difficulty of image segmentation is always determined by the extent of image pre-processing. In image segmentation, one document image is usually classified as four types: background, text, graph, and photograph since data types of document images always depend on the user's definitions. But how to find the features of each data type is still very difficult and irregular. And it will much depend on the complexity of the original images. So, image documentation is one of the main topics of image processing in our group all the time. We will show some examples of original and processed images in the below figures.

3. Document Processing:

 Since there are many research subjects in image documentation, we can always make some modifications about the document images for further processing, including angle detection, the calibration of the skewed images, and the morphological changes of the document images. We will show one of the results of all these aspects such like the calibration of a skewed document image in particular of printed images.

4. Image Post-processing:

 To reach a higher end we usually have to apply more processing steps after our first processing. We have numerous methods such like context-based, boundary refinement, multi-channel skills, and so on, to improve our final experimental results. And image post-processing also plays an essential role in image documentation beside of our common treats to documentation.

中 文 說 明
隨著資訊媒體快速的發展,文件影像處理的討論已有漸漸萌芽及發展的趨勢;普遍所及,許多的影像文件處理問題早已充斥於我們生活中了,不論是掃瞄、列印等各種電腦相關應用技術都和我們的日常生活有著密不可分的關係,基於種種考量及以上原因,我們這個研究團隊特別針對如何使影像文件呈現出更好效果以專心致力於這項研究議題,以下我們就簡單列出我們的幾項研究方向:

1. 文件影像之前處理:  由於一般的文件影像處理所會遭遇到的問題往往沒辦法直觀的找出其解決方法,除此之外,為了使處理的問題影像更加的單純,在文件影像處理中的前處理就格外的重要,其目的主要是為了讓原始影像的雜訊能夠大大的減少,使得之後的文件影像處理能變的更加的容易,所以這裡,我們特別列出幾項在前處理中重要的研究議題,這些問題也是我們所要解決及研究的方向之一,如de-shading, auto-cropping, de-bleeding等等…

2. 圖文分離:  如前所述,文件影像的前處理做的好壞會直接影響到之後分析處理之難易,這也就是為什麼圖文分離的難易度和效果好壞會和影像前處理的結果優劣有著密切的關係,對圖文分離而言,一張文件影像基本上可以分成四種資料型態:分別為背景、文字、圖片、及照片等。而文件影像要被分類成的資料型態數目都是取決於使用者的自行定義,但是如何找出每種資料型態的特徵不僅困難而且毫無規則性可言,當然這也和原始影像的複雜度有極大的關連,因此,文件影像處理一直在我們研究團對裡扮演著一個相當重要的角色,也是我們主要的研究課題,下面的圖片是特別對圖文分離的應用例子做一大致的圖略說明。

3. 文件影像處理:  因為在文件影像處理當中存在有許許多多的研究主題,所以我們往往會對文件影像的進一步處理做些許的修正,好比文件影像的角度偵測、傾斜角度校正、還有文件影像的型態學變化分析等等…下面我們也特別對其中某些方面列出我們的一些研究結果,如在特定列印影像中傾斜文件影像的校正。

4. 文件影像之後處理:  為了使文件影像的處理效果達到更佳之目的,我們通常會在我們第一次的處理之後,做些額外的處理步驟,針對這個目的而言,目前學術界存在有許多的方法來解決此一問題,一般較常使用的像是以處理區塊的周圍資訊來判斷、邊界限制的使用、多次通道處理技術等等…用以改善最後的實驗結果,除此之外,文件影像後處理常使用的技術在文件影像分析中也扮演著一個不可或缺的角色。