Design of irregular screen sets that generate maximally smooth halftone patterns

A Study on Halftoning Improvement for Low-Resolution Digital Print Engines With Machine Learning Methods

“…It shows that the high noise levels of an irregular screen originate from the presence of low (quantization) frequencies that are absent from the spectrum of regular screens. In [12] it is suggested to quantitatively assess the additional noise of an irregular screen by summing the energy of the quantization frequencies in Fourier space and normalizing the energy in the frequencies associated with it, with a corresponding regular screen. In this method, each irregular screen receives a grade (ratio) that is used to assess the expected noise of the screen in the print.…”

“…These additional frequencies cause quantization patterns to appear in the print. The common practice to incorporate irregular screens in the screen design is to calculate these patterns directly from the reciprocal screen lattice (Fourier representation of the screen lattice vectors ( 11)) followed by a threshold, rule-based operation that determines if the screen is in the acceptable or unacceptable domain [11], [12], [13], [8], [10].…”

“…Next, for each screen, the 6-dimensional quantization vector (6 lowest and strongest quantization frequencies) was calculated via (17). To determine the screen class, we followed the legacy rule-based approach [8], [12], [10]: if the six lowest quantization frequencies of a given screen were higher than Indigo's empirical thresholds (Table 4), the screen was tagged as class 1 (acceptable); otherwise, it was tagged as class 0 (unacceptable). Two types of vector features were used to construct the data set.…”

“…Thus, if the resolution is too low, as in the case of digital print engines, it is usually impossible to design a high-quality, moiré-free screen set [11]. One solution to this problem is the method of irregular screens [11], [12], [13]. Irregular screens form a space-varying cluster grid in which 1/distances and the orientation of two adjacent clusters vary across the grid and are equal on average to the screen resolution and angle, respectively.…”

“…When the patterns are too strong, the screen is considered unacceptable for production. The classical rule-based approaches identify quantization-free screens by performing Fourier analysis of the screen spectrum and avoiding screens that contain low frequencies in the screen design process [12], [13]. These approaches use overly simplistic models to simulate the printing process and the human visual system (HVS).…”