A caesium magnetometer attached to a handcart enables accurate magnetic prospection of large areas within short times. The roughness of the earth’s surface, minor differences in handling the handcart and a lot of other reasons are possible sources for various errors during a survey. Standard image preprocessing methods, such as median or low‐pass filtering, remove not only errors but also important information and deform the shape of magnetic anomalies. We present four special preprocessing methods to remove errors automatically by protecting and restoring the real value and shape of anomalies and compare them with standard preprocessing methods. The four methods are equalizing subgrid shifts, equalizing line shifts, removing spikes and removing displacement errors.