A method of using a charged particle microscope comprising a source; a specimen holder, for holding a specimen; an illuminator, for irradiating the specimen; a detector; and a controller, for controlling at least some aspects of the microscope's operation. The method comprises the steps of using the detector to acquire a series of component images of a part of the specimen; then successively quantizing each component image and storing it in a memory; recording a quantization error per pixel for each quantized component image, and keeping a running tally of cumulative quantization errors per pixel for the quantized component images; when quantizing a next component image, choosing a quantization polarity for each pixel that will avoid further increasing the total quantization error for each pixel. Finally, combining the component images to assemble a composite image.