影像處理之網紋分析及其應用

Abstract

當掃描印刷品文件時，常常會在掃描結果中發現網紋雜訊，這種雜訊不僅影響掃描的品質而且使得文件分析及影像處理的過程複雜化。本論文使用符立葉分析，以解釋半色調網點的產生及掃描網點後形成網紋現象的成因。 為了分析半色調的網點及其被掃描後所產生的網紋，網點參數的度量尤為重要。本論文中提出一個數位網點的度量方式，它能夠在高解析度半色調網點影像頻譜中，利用各種網點的特性自動地度量出網點的角度與頻率。 利用網紋分析的結果及對各種網紋現象的掌握，本論文提出了各式抑制網紋的掃描策略。其中包含雙掃描行程及網紋控制去除法。雙掃描行程是一種可以利用硬體實作的快速網紋抑制掃描術，將相同解析度但是具水平或垂直偏移掃描格的兩次掃描結果平均，便可有效去除主要的網紋。網紋控制去除法是利用符立葉分析推導出一組公式，以便於計算稱為網紋控制解析度的值。利用這個掃描解析度掃描出來的影像，出現網紋的訊號會收斂到特定的頻率空間中，然後利用事先設計好的濾波器除去這些頻率的訊號，達到抑制網紋的效果。本論文中并提出數種濾波器以供不同的狀況下使用。 本論文也提出數種利用網紋分析及控制技術的應用。其中包含利用網紋分析及控制技術，完成文件分析及文件品質改良的工作；以及利用利用網紋的混合達到印刷品商標隱藏的效果。由於網紋的出現是因為對規律的訊號做偏移取樣的結果，而在印刷品上又可常常發現各種半色調影像、色塊或字行等會出現規律網點或是文字的圖案，這些規則性訊號經過取樣必然會產生網紋，而這些網紋可被用來做為文件影像切割依據。首先利用掃描機解析度的設定控制掃描出來網紋的頻率，然後利用特殊設計的梳型濾波器偵測出印刷品半色調網點以及文件段落等規律訊號出現的區域，最後利用網紋控制去除法去除網紋雜訊，達到文件分析及文件品質改良的效果。此方法可以有效的區隔出印刷文件上文字區塊、色塊、圖案、以及半色調影像的區域，并且可同時消除印刷網紋改善掃描影像的品質。相類似的網紋控制技術也可反向用來製造特殊的半色調影像並隱藏商標或是版權申明圖案於有價值的印刷品中。本論文中提出在即將印出的半色調影像中選擇出商標或是版權記號形狀的區域並利用不同的網點成相；利用半色調的高頻率結構隱藏這些圖案，使得肉眼難以分辨這些被隱藏的圖案；再利用不同的半色調網紋印刷品會在重新掃描時產生不同的網紋的特性，使得印刷品一經掃描後便出現這些商標或是版權記號。本論文中也提出一種網紋能見度圖表，可用來選擇不同的印刷網點，使得成相出來的半色調網點結構會在掃描結果中呈現較大的網紋差異。同時一種六角形網點也被提出，用來提昇這種印刷術的穩定度。

Moire patterns are distortions on the results of scanning printed documents. It degrades the resulting image and makes document analysis more difficult. The Fourier analysis can be employed to describe the moire phenomenon. In this dissertation, the Fourier analyses of both the screening and the scanning processes are presented, from which the moire patterns can be modeled. The screen parameters like the screen angle and frequency are necessary for the analysis of halftone screens. A digital screen tester is proposed which can be used to measure the screen angle and frequency. By some criteria of the halftone screens, the tester determines automatically the screen angle and frequency by analyzing the Fourier spectrum of a high-resolution portion of the printed picture. By using the knowledge of the moire patterns, strategies to suppress the moire patterns in the scanned halftone images are proposed. A double-scan process, which can be implemented by hardware, is introduced first. After the first scan of a halftone image, the scan position is shifted by half of the sampling grid distance in either or both of the horizontal and vertical directions. Then by averaging the images of the two scans, most of the moire fringes can be removed. Also, based on the Fourier analysis, a formula is derived to calculate a special scanning resolution, called moire controlling scanning resolution. With this scanning resolution, an input halftone printing is scanned and the moire signals can be confined in programmed frequency areas. After filtering out the signals in these areas, the original can be restored. Some useful spatial filters for this purpose are designed for different cases of moire signal removing. Furthermore, the technique of moire signal controlling is utilized in document image segmentation and document quality improvement. Since the moire phenomenon comes from sampling periodical structures, scanning periodical signals such as halftone screens, color components, and text galleys in printed magazines and newspapers will generate moire patterns. The moire patterns can be utilized as features for document segmentation. By using the moire controlling scanning resolution, moire patterns with certain frequencies and angles are generated on the scanned document image. Then, a logical filter, called the comb filter, is proposed to detect the moire pattern and to identify the location of the periodical components of the document. The proposed method can extract effectively gray or color pictures, artworks, and text blocks in printed documents. Moreover, the moire patterns on the segmented document components can be suppressed easily. The suppression yields better image quality for further analysis and image compression. Finally, a new method to embed copyright patterns into printed halftone images is proposed. The proposed copyright-labeling method produces a halftone image from the original image by mixing halftone screens. A copyright pattern, such as a company logo or a trademark, can be included as an image region in which different halftone screen is applied. Because the frequencies of the halftone screens are much higher than the human detectable bandwidth, the hidden copyright patterns can be hardly identified by human eyes. However, the periodical high frequency signals are sensitive for scanners. One who reproduces the image by scanning will find that the copyright patterns are generated on the scanning result in the form of moire patterns. A moire visibility map is proposed to help selecting pairs of halftone screens for mixing the copyright pattern and the halftone. As a result, on the re-scanned image the corresponding moire patterns of the two halftone screens will have larger diversity and are easier to be distinguished. Also, a hexagonal screening grid is proposed to improve the consistency of the screen mixing. Experiments have been conducted and good results prove the feasibility the proposed approaches. It is worth to develop more applications of the proposed methods.