平 成6年 度 照明 学会全 国大 会 IS-07 TESTING COLOUR MODELS PERFORMANCE USING COMPLEX IMAGES M.C. Lo (Loughborough University), M.R. Luo, and P.A. Rhodes (University of Derby) INTRODUCTION The work described here closely follows the CIE TC1-27 guidelines1 on the evaluation of colour models for reflection print and self-luminous monitor image comparison. Nine models are studied and these can be divided into three categories: the CIE colorimetric system2 (XYZ, CIELAB, and. CIELUV), colour appearance models (Nayatani et a13, Hunt4, Fairchild and Berns5 cognitive and sensory models), and chromatic-adaptation transforms (von Kris6 and BFD7). At present, experiments are in progress and the results will be presented at the 1994 Annual IEI-J Conference. IMAGE PREPARATION Both reflection print (hardcopy) and self-luminous (softcopy) images are used in the experiment. Eight different transparencies were selected including man-made objects, people , and natural scenes. These were input-using a Crosfield scanner to obtain CMYK digital data. These data were used to print the Cromalin hardcopies used in the experiment. For softcopy images, the data was processed in three stages to: 1) convert CMYK in terms of tristimulus values (under hardcopy viewing conditions) using a black printer algorithm developed earlier8, 2) obtain the corresponding tristimulus values (under monitor viewing conditions) using a particular colour model, and 3) transform the corresponding tristimulus values into RGB signals9. A white border was added to each image as an anchoring point (reference white) . EXPERIMENTAL CONDITIONS The experimental setup involves a hardcopy image illuminated in a Verivide viewing cabinet and a pair of images with the same content displayed on a Barco monitor. Fig. 1 illustrates the experimental situation with a pair of softcopy images on the right, and a hardcopy image on the left. The same physicalsize is used for both the hardcopy and softcopy images . On the i monitor, a set of buttonsis used for scaling quality of reproduction. These comprise numbers from one (for exact match) to seven (for bad match) . All images are surrounded by a medium grey with an L* of 50 relative to the Cromalin substrate and the white point of monitor. The whole experiment is divided into two phases according to the light sources used in the viewing cabinet and the white points on the monitor. In phase one, the chromaticities of white points for monitor and the light sources for the reflection hardcopy are the same. A D65 and a D50 simulator are used in this study. In phase two, the chromaticities of the white point for the monitor and those of the light sources for the reflection hardcopy differ . Three sets of viewing conditions are studied: Light source White point (viewing cabinet) (monitor) i) D50 D65 ii) D50 9300K iii) D65 9300K The luminance level for the three white points studied is set at 70 cd/m2 . The luminance value of the hardcopy's border is adjusted to instrumentally match that of the softcopy's border by altering an intensity controller in the viewing cabinet. ―88―