Non-linear filters for removing noise from binary images

The standard strategy described is highly effective for improving detection performance in many practical problems. Some variations are superior to the binary median filter which is frequently recommended. Scope exists nevertheless for developing the approach, for example, by storing the original triggers until filtering is complete so that degraded trigger clusters may be restored and the `W´ problem is overcome. The author shows that the SA (simple association) filter may be improved by implementing it as a succession of stages, with small blocks examined first when they are then considered as examination cells for larger blocks. If the binary image is held in framestores and processing time is not a limitation, the range of approaches possible becomes very wide