Grey-scale 1-d dilations with spatially-variant structuring elements in linear time

Spatially variant morphological operators can significantly improve filtering capabilities or object detection score of various applications. Whereas an effort has been made to define the theoretical background, the efficient implementation of adaptable algorithms remained far less considered. In this paper, we present an efficient, one-scan, linear algorithm for 1-D grey-scale dilations/erosions with spatially variant structuring elements. The proposed algorithm processes data in stream, can work in place and produces results with minimal latency. The computing time is independent of the structuring element size.