Banding measurement for color laser printers color management system for digital cinema glare and shadow reduction

Thanh Huy Ha, Purdue University

Abstract

In this dissertation, we first introduce a new banding analysis system which contains a banding measurement tool and a spatial-visualization tool. The banding measurement tool is based on image analysis to measure color banding exhibited by color laser printers. The tool analyzes a specially designed test page and provides the principal banding frequency components from both the primary and secondary colorant planes. On the other hand, the spatial-visualization tool is intended to help the users better understand from a qualitative perspective the appearance of the color banding on their printed test pages. The overall banding analysis system is very useful for engineers in printer industry to investigate banding in color laser printers throughout all stages of the printer development life cycle. In the second part, we describe an accurate color management system (denoted as NLXWP) to perceptually map between two different monitors. The ultimate objective is to create a color-consistent environment for applications in the digital cinema. We introduce two new tools specially for the development of the NLXWP model: a white point correction method, and a highly accurate monitor characterization approach using multiple non-square matrices. The white point correction for each monitor is accomplished by adjusting the balance between three color channels of the monitor. Additionally, we provide a tolerance parameter to control the trade-off between the accuracy of white point correction and the preservation of monitor peak luminance. Besides, our new monitor characterization approach employs a color classifier followed by four non-square transformation matrices in order to achieve better accuracy than that of other previously reported model-based methods. We also discuss a discontinuity issue on the boundaries between the color classes, and propose two smoothing techniques, weighted-sum and overlapping training, to handle it. The experimental results show that the NLXWP model works quite well to perceptually map between two particular target LCD monitors. Finally, we describe an algorithm for mitigating glare and shadow in images taken by a camera system. The system contains a digital camera and three Light Emitting Diodes (LED) located at center, right, and left of the camera. For each object to be taken, the system will capture it in four times. During the three first captures, the three LEDs (center, right, left) take turn to illuminate the object. During the fourth capture, all LEDs are turned off and the object is captured under the ambient light. The mission of our algorithm is to combine information from the four frames obtained with the system and generate an image in which glare and shadow are mitigated. To achieve this objective, our algorithm first detects shadow in each frame then multi-resolution-fuses the three first frames corresponding to the center, right, and left LEDs while excluding the shadow pixels detected in the first step. In the fusion process, we also set very low weights for pixels that have too high luminance values (near 255 in the case the camera uses 8-bit-integer for each color channel). These bright pixels are supposed to be too saturated and suspected to be glares. As a result in the final fused image, both glare and shadow are significantly mitigated.

Degree

Ph.D.

Advisors

Allebach, Purdue University.

Subject Area

Computer Engineering

COinS