Halftones and other binary images are difficult to process with causing sev
eral degradation. Degradation is greatly reduced if the halftone is inverse
halftoned (converted to grayscale) before scaling, sharpening, rotating, o
r other processing. For error diffused halftones, we present 1) a fast inve
rse halftoning algorithm and 2) a new multiscale gradient estimator, The in
verse halftoning algorithm is based on anisotropic diffusion. It uses the n
ew multiscale gradient estimator to vary the tradeoff between spatial resol
ution and grayscale resolution at each pixel to obtain a sharp image with a
low perceived noise level, Because the algorithm requires fewer than 300 a
rithmetic operations per pixel and processes 7 x 7 neighborhoods of halfton
e pixels, it is well suited for implementation in VLSI and embedded softwar
e. We compare the implementation cost, peak signal-to-noise ratio, and visu
al quality with other inverse halftoning algorithms.