Clustered-dot halftoning with direct binary search
Most printers based on electrophotographic technology use periodic, clustered-dot screening for rendering smooth stable prints. Despite great advancement in the print quality from periodic clustering, there exists one major drawback, namely periodic moiré, which is caused due to interference between multiple periodic screens superposed for color printing. Researchers have proposed solutions to this problem in the form of optimized screen periodicities, screen angles, and screen growth sequences. These solutions work well, but only when the involved color planes are few in number. A common scenario nowadays is the use of six or more primary colors. An alternative solution to avoid periodic interference between conventional clustered-dot screens is to use stochastic clustered-dot screens. A set of stochastic screens can be designed with optimum coarsenesses to be used exclusively or in conjunction with periodic screens to greatly reduce the chances of undesirable interference. In this thesis, we propose a method of stochastic, clustered-dot halftoning and screen design based on Direct Binary Search (DBS). In earlier publications, DBS was shown to generate stochastic, dispersed-dot texture. An analysis of the relationship between halftone changes and the update of filtered-error in DBS lays ground for the development of CLU-DBS which originates as a clustered-dot halftoning variant of DBS. The CLU-DBS procedure evolves from DBS by a simple modification of the filter sizes. CLU-DBS uses the optimization framework of DBS. Like DBS, it first initializes the filtered-error LUT on the basis of an initial random halftone. Subsequently, it iteratively selects pixels for toggle and swap with an objective to minimize the magnitude of the filtered error, and updates the filtered-error. The essential deviation from DBS is the use of different filters in the initialization and update steps. CLU-DBS uses an update filter of larger size than the initialization filter. After proposing an intuitive explanation for the development of stochastic, clustered-dot texture from CLU-DBS optimization, we derive a closed-form expression for the underlying cost function that is being minimized in the course of optimization. Later this cost function is established as the guiding metric for stochastic, clustered-dot halftoning. In other words, CLU-DBS is implemented based on the cost metric. The dual-component cost metric reveals a great deal of information about the relationship between the input parameters and the output texture. We make use of this information to improve the quality of output texture. This is done by an intelligent selection of input filter sizes, the initial halftone, and the optimization procedure. After developing a procedure for halftoning an image with CLU-DBS, we develop an algorithm for screen design. The screen design is complete after being post-processed for removal of certain defects from the tone reproduction curve. In order to evaluate the quality of halftones resulting from designed screen, we performed psychophysical experiments. The experiments involved comparison of CLU-DBS printed halftones with the halftones resulting from methods developed previously. The results of these experiments indicate subjects' clear preference of the CLU-DBS halftones over other halftones in terms of smoothness of prints.
Allebach, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our