The error diffusion algorithm is widely used in digital image halftone. The algorithm is very simple to implement and very fast to calculate. However, it is known that standard error diffusion algorithms such as Floyd-Steinberg produce undesirable artifacts in the form of structure artifacts such as checkerboard patterns, diagonal stripes and other repetitive structures. The boundaries between structure artifacts break the visual continuity in regions of low intensity gradients and may be therefore responsible for false contours. In this paper, we propose a new halftone method to reduce the structural artifacts and to increase the gray expression, called hybrid error diffusion, by the concept of “error diffusion by perturbing the error coefficient with randomly selected mask”. The proposed algorithm consists of two steps in each position. In the first step, a perturbation is calculated using the internal pseudorandom number and randomly selected 4x4 mask, similar to dither mask (R,G,B different). In the second step, error diffusion weights are calculated with the criterion for each pixel values. So, the proposed hybrid method can reduce the structure artifacts with keeping the advantage of the error diffusion. This paper discusses the performances of proposed algorithm with experimental results for natural test images.