Objective
Digital museums break through the barriers of space and time, expanding the range of collections in the form of video and images. However, in the process of museum relic image collection or online display, image reproduction across equipment is prone to color distortion phenomena such as eclipse, color deviation, affecting the visual effect of the image due to the correlation characteristics of color equipment and the lack of standard image collection conditions. In order to solve this problem, the Qianlong palette silk fabric was taken as the research objective to carry out the cross-device color reproduction of textile relic images, aiming to provide support for the establishment of digital museum.
Method
On the basis of Qianlong palette silk fabric, the color characterization of digital camera and display screen were carried out respectively. The influences of different image formats, different training data sets and different order of polynomial on the accuracy of camera prediction model were discussed. At the same time, the Gain-Offset-Gamma(GOG) model was used to characterize the chrominance of display screen. Then color difference analysis was carried out to determine the optimal characteristic model of Qianlong chromatography reappearance across equipment. Finally, textile relics collected in the Silk Museum of China were used to test the model.
Results
In this study, color information and image information were collected from 210 colors of Qianlong palette. The absence of Qianlong palette in blues (B) and blue-green tones (BG) may be due to the fact that indigo was the only natural blue dye selected in this study, and a stronger dye bath may have biased indigo stained samples towards purple tones (Fig. 6). The L* values of all 210 colors ranges from 17.85 to 78.13(Fig. 7), 210 colors are distributed in four quadrants of the a*-b* plane of CIELAB color space, and is widely distributed in the first quadrant (Fig. 8). The Qianlong palette covers different color series. Among all the colors, red and yellow are the main color series, which are the most important color representations of the royal costume in the Qing Dynasty. The color reproduction accuracy for the silk fabrics has been significantly influenced by the training datasets and the mathematical mapping methods and image format. Using both color charts, the RAW image format shows the better predictive accuracy, followed by the JPG format (Tab. 6), and the 3rd polynomial regression shows the best predictive accuracy (3.20 (ΔEab*)), followed by the 2nd (3.86 (ΔEab*)) and the 1st order polynomial regression (4.15 (ΔEab*)). Finally, the RAW image and third-order polynomial were selected to characterize the chrominance of the camera, and the collected textile color data set was used to train the model. The accuracy of the model can reach 2.18(ΔEab*), this is the highest precision camera characterization model in this study. The accuracy of display characterization model is 0.58 (ΔEab*). The mean color difference value of Qianlong palette and silk relics is 2.16 (ΔEab*), and 2.38 (ΔEab*), indicating that this characteristic model has excellent cross-equipment color reproduction effect.
Conclusion
The Qianlong palette is mainly distributed in the yellow tones of the Munsell color system, with less color in the green and purple tones, and no distribution in the blues and blue-green tones. In the color space of L*a*b*, the brightness value of Qianlong chromatographic ranges from 17.85 to 78.13. The brightness value of yellow tone is above medium, and the brightness value of other tones is below medium. For museum image acquisition, it is suggested to standardize the lighting source, which mainly includes the selection of standard lighting body and corresponding color temperature to ensure the uniformity of lighting. At the same time, standardize shooting conditions: select and fix appropriate camera parameter settings, shooting distance, choose remote control shooting to avoid human operation error. For the characteristic prediction model of camera: the textile color data set of the same material and the same specification is used as the training data set, and the image modeling in high order polynomial and RAW format is adopted to obtain higher prediction accuracy and accurately predict the color after textile reduction.
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)