WInHD: Wavelet-based Inverse Halftoning via Deconvolution

Ramesh Neelamani, Robert Nowak, Richard Baraniuk

(Submitted October 2002)

We propose the Wavelet-based Inverse Halftoning via Deconvolution (WInHD) algorithm to perform inverse halftoning of error-diffused halftones. WInHD is motivated by our realization that inverse halftoning can be formulated as a deconvolution problem under Kite et al.'s linear approximation model for error diffusion halftoning. Under the linear model, the error-diffused halftone comprises the gray-scale image blurred by a convolution operator and colored noise; the convolution operator and noise coloring are determined by the error diffusion technique. WInHD performs inverse halftoning by first inverting the model-specified convolution operator and then attenuating the residual noise using scalar wavelet-domain shrinkage. Since WInHD is model-based, it is easily adapted to different error diffusion halftoning techniques. Using simulations, we verify that WInHD is competitive with state-of-the-art inverse halftoning techniques in the mean-squared-error sense and that it also provides good visual performance. We also derive and analyze bounds on WInHD's mean-squared-error performance as the image resolution increases.

Support: This work was supported by the NSF grants CCR--99--73188 and MIP--9701692, AFOSR grant F49620--01--1--0378, ONR grants N00014-02-1-0353 and N00014--00--1--0390, ARO grant DAAD19--99--1--0290, and Texas Instruments.