以視覺評量法探討數位影像知覺品質之研究

A Study on Perceptual Digital Image Quality by Visual Assessment Method

液晶電視已然成為家電市場的必要趨勢，其廣泛的應用也吸引數位影像與平面設計等專業工作者，使之成為創作作品的展示平台。因此，隨之衍生的影像顯示品質議題，亦成為影像與顯示科技產業研究者所關切的重點工作。本研究主要目的，在於透過視覺評量實驗法，探討人們對於液晶電視所知覺的影像品質。實驗採用之影像樣本，為日常生活常見之七大類影像。透過三位設計與色彩領域專家進行焦點小組討論，並針對各類別的影像依據：紅、綠、藍、黃，以及無彩色等色系為影像的挑選基礎。挑選後的35 張影像，則依據影像物理屬性，包含：影像對比度、影像飽和度、影像銳利度等進行實驗刺激材料之準備，並以便利抽樣方式，針對7 位具有經常操作「影像處理」經驗的受測者，進行影像品質之視覺評量實驗。實驗結果發現，對比度較高的影像，人們所知覺的影像品質相對優於對比度較低的影像；高飽和度的影像，其影像的知覺品質相對優於低飽和度的影像。此外，評量結果發現，影像銳利度參數與影像銳利度知覺品質評量結果，呈「倒V」的非線性關係，且影像銳利度參數設定於「+6」與「+8」時，其影像品質明顯優於「+0」、「+18」與「+20」的參數。本研究同時以典型相關分析，進一步探討「影像物理屬性」對「影像知覺品質」之影響。就整體而言，模式一的結果顯示，影像物理屬性有57.94%可以解釋影像知覺品質；模式二的結果顯示，影像物理屬性有47.36%可以解釋影像知覺品質。因此，影像對比度對於影像知覺品質的影響程度，高於影像飽和度與影像銳利度。本研究所探討「影像物理屬性」與「影像知覺品質」之間的關係已獲得具體的成果，該結果可提供影像設計工作者或液晶電視影像顯示驅動晶元設計與相關研究單位之參考。

Not only has the LCD-TV become more popular in our life, but it also has been used to display digital image and graphic design by artists and designers. Thus, researchers of the image and display technology industry regard the image quality as an important issue in recent years. Therefore, this study focuses on the perceptual image quality of LCD-TV by visual assessment method. Seven types of image samples were set for the experiment, each of which included four images of primary color （red, green, blue and yellow） and one image of less saturation. Finally, 35 images were used in the experiment of image quality judgment. Each of these images had been systematically controlled in contrast, saturation and sharpness effects by using ‘Adobe Photoshop' software. For example, in terms of contrast effect, images ranged from original image （zero） to positive twenty levels and from original image to minus twenty levels each of which differs in 2 levels. Seven subjects' data of visual judgment were collected for further analysis. Results of the ANOVA showed a significant main effect of ‘visual judgment' without respect to contrast and saturation of image. When the image contrast is higher, the image quality is better than others, and when the saturation is higher, the image quality is better than others. In addition, the correlation between sharpness of image and image quality is ‘Nonlinear' as like as an ‘invert-V'. The quality of image samples from the positive sharpness levels 6 and 8 are significantly better than levels 0, 18, and 20. Results of the Canonical Correlation Analysis showed that physical attribute could explain 57.94% of the perceived image quality in model 1, and 47.36% in model 2. So, the effect of contrast or saturation of image is higher than that of sharpness of image in the influence of perceived image quality. In conclusion, this study explored the relationships between physical attribute （e.g., contrast and saturation of image） and perceived image quality. It is expected that the results obtained from the study can provide engineers and designers of the LCD-TV color chips with a better understanding of humans' subjective perceptions on image quality.