Sigmoid函數對修正色彩分佈指標影響之研究

Abstract

色彩再現需要影像品質評量，過去學者提出以最小可視覺 (Just Noticeable Difference, JND)與色彩分佈指標(Color Distribution Index, CDI)，以人類對色彩與亮度來評估人類視覺的色彩再現。由於描述此理論所使用之CIE1976 CIELUV色彩空間，其色彩空間在亮度低於100 cd/m2 時呈現增加的趨勢，此現象違背人類於低亮度的實際感知。因此本論文提出名為修正色彩分佈指標影響(modified Color Distribution Index, mCDI)之定義，以CIELAB色彩空間取代CIE1976 CIELUV色彩空間。此外，本論文提出以Sigmoid函數取代傳統所使用之gamma函數作為灰階調整參數，經驗證只需要11-bit色深達成100% 之色彩再現，而傳統之gamma函數需要12-bit色深達成。在本論文中亦分別以色彩分佈指標對色深、色彩範圍、標準白點、最大亮度與對比度進行討論。

Image quality evaluation is required for color reproduction. Just Noticeable Difference (JND), and Color Distribution Index (CDI) were proposed for evaluating the reproduction of human visual gradation. Since the description of CDI uses the CIE1976 CIELUV color space. The gamut size in CIELUV color space increases as the luminance decreases from 100 cd/m2 doesn’t correspond to human perception at low luminance. In this research, we proposed modified Color Distribution Index (mCDI) applies in CIELAB color space, instead of CIE1976 CIELUV uniform color space, approximate human vision and uniform color space. We also proposed that sigmoid function is used for gray level modulation, instead of gamma function. We had clarified that 11 bits of the sigmoid function is needed for approaching 100% color reproduction based on BT.709, and 12 bits is needed for gamma correction. Also, we discussed the effects of bit depth, gamut size, white point, maximum luminance, contrast ratio in CDI evaluation.