科技部專題研究計畫主持人
108-2410-H-224-021-
以演算法解決造字灰度控制之研究
The Study on the Grey Scale Control in Chinese Font Design with Algorithm
科技部
Ministry of Science and Technology
國立雲林科技大學視覺傳達設計系暨研究所
Department of Visual Communication Design, YunTech
胡文淵
WEN-YUAN, HU
國立雲林科技大學視覺傳達設計系暨研究所
Department of Visual Communication Design, YunTech
胡文淵
副教授
0988036511
iuan.hu@msa.hinet.net
計畫執行期間起:2019-08-01
計畫執行期間迄:2020-07-31
2019-08-01
2020-07-24
文鼎晶熙黑體_M、文鼎明體_M、文鼎仿宋體_M、文鼎楷體_M文字灰度,每種字型根據《國小學童常用字詞調查報告書》全字5021字。
3
096
07
像素密度;演算法;點陣灰度;字面;筆畫
Pixel Density;Algorithm;Lattice Grayscale;Area of Characters;Stroke
本研究之目的是探討預測中文字體灰度的有效方法,從前導性研究中已得知,字面大小的主觀感知會受到文字幾何構形的視覺影響,該實驗結果顯示凸形封包作為文字的字面,在研究分析上比矩形字面更能貼近真實。因此,在此研究基礎上,本研究選定「連通元件對文字凸形封包的像素比值」作為和筆畫數搭配的一組預測變數,並就文獻所提示的字符總密度值和點陣灰度值兩個依變項進行相關性分析,以判斷建立預測模型的可行性。 為建立一個有效的預測模型,除了變異項要有足夠的相關性和解釋力,本研究認為數據量是決定預測準確度的關鍵。故研究取材選定文鼎晶熙黑體及文鼎明體、文鼎仿宋體、文鼎楷體等四個基本印刷字型產品,並依據教育部頒布之《國小學童常用字詞調查報告書》字頻總表,共5,021字為文字取樣範圍。循著文獻提示的字符總密度公式及點陣灰度公式,設計演算程式並進行文字的灰度演算。 研究結果顯示:(1) 字符總密度公式無法運算獨體字灰度是無法克服的應用侷限,且字符總密度公式的某些組成算式會在演算時受到筆畫重疊的影響而產生數據錯誤,兩項因素及相關性檢定判斷字符總密度公式無法建立有效預測模型。(2) 點陣灰度值則能在筆畫數與「連通元件對文字凸形封包的像素比值」做為預測變數的條件下,能有效建立預測模型。(3) 黑體、明體、仿宋體、楷體之可依平均點陣灰度值的比值關係0.32072924:0.27311465:0.19527939:0.20907024得到系統性的灰度轉換,在四個受測字型中以黑體及明體的預測效果最佳,分別達到 92.4%和 92.7%的解釋力。根據上述結果,筆畫數與「連通元件對文字凸形封包的像素比值」確實可以建立預測點陣灰度的有效模型。
The purpose of this study is to explore effective methods to predict the grayscale of Chinese fonts. It has been known from the pilot study that the subjective perception of the size of the Area of Characters will be affected by the visual effect of the geometric configuration of the text. The experimental results show that Convex Hull, as the Area of Characters of text, is closer to reality than the rectangular Area of Characters in research and analysis. Therefore, under such a research basis, this study selected “Connected Components Area Ratio (CCAR)” as a set of predictive variables in combination with number of strokes. In addition, this study performed a correlation analysis on the two dependent variables, the total character density value and the lattice grayscale value suggested in the literature to determine the feasibility of establishing a prediction model. In order to establish an effective prediction model, in addition to the variation items that have sufficient relevance and explanatory power, this study believes that the amount of data is the key to determining the accuracy of the prediction. Therefore, this study selected 4 basic printed font products, including Arphic Jingxi Bold, Arphic Ming, Arphic Imitating Song, and Arphic regular, as the research materials. Moreover, based on the Survey Report on Commonly Used Words by Elementary School Children issued by the Ministry of Education, a total of 5,021 words are selected as the text sampling range. This study used the character total density formula and dot matrix grayscale formula suggested in the literature to design the calculation program and perform the grayscale calculation of the text. The research results show that: (1) The total density formula of the characters cannot calculate the grayscale of the single font, which is an insurmountable application limitation. Moreover, some of the composition formulas of the total character density formula will be affected by the overlap of strokes during the calculation, resulting in data errors. The two factors and the correlation test determined that the total character density formula cannot establish an effective prediction model. (2) The lattice grayscale value can effectively establish a prediction model under the condition that number of strokes and “Connected Components Area Ratio (CCAR)” are used as predictive variable. (3) Based on the ratio relationship of the average lattice grayscale value 0.32072924: 0.27311465: 0.19527939: 0.20907024, Arphic Jingxi Bold, Arphic Ming, Arphic Imitating Song, and Arphic regular cano obtain a systematic grayscale conversion. Among the 4 tested fonts, Arphic Jingxi Bold and Arphic Ming showed the best predictive effect, reaching 92.4% and 92.7% of explanatory power, respectively. According to the above results, number of strokes and “Connected Components Area Ratio (CCAR)” can indeed be used to establish an effective model for predicting lattice grayscale.
否,無直接識別資料
除檢查有無可識別個人身份資料外,另檢查個案編號是否重複、變項及選項數值說明是否與問卷一致或未被定義,發現前述狀況時,將於確認後協助修改。
追蹤清單
下載(0)
申請(0)
遠距(0)