Qualitative comparison of visual models in an iterative halftoning procedure

Visual models are increasingly being incorporated in digital halftoning algorithms. In order to select the most appropriate visual model, it is important to gain an understanding of the visual effects of the different visual models on the halftone output. This work compares the effects of two single-channel visual models on halftone output by using an iterative technique that explicitly shapes the quantization error in the Fourier domain. By operating on the spectrum of the quantization error, we are able to directly apply a frequency sensitivity visual model as an error-spectrum shaping function. Our results clearly indicate that a low-pass visual model produces better quality halftones than a conventional contrast sensitivity function model. In addition, we introduce an image adaptive method for shaping the spectrum in a way that preserves perceptually important high frequencies without sacrificing the desirable effects of visual model based constraints