We developed a newhalftoning method, adaptive cell, which generates unstructured dots as by error diffusion method, but the produced binary image is less granular and more stable than that of error diffusion. The basic idea is covering the input image with cells which vary in size and shape depending on the input image data. Then appropriate number of pixels at the center of each cell are set on, to preserve the local average of the input image. The characteristics of the generated binary image shows that this method is suitable for high density electro photographic binary printers.