Separable 2-D filters for the detection of ramp edges

The author examines the use of 1-D optimal convolution filters for the detection of ramp edges. The idea is to replace one 2-D convolution by two 1-D ones performed in sequence. The way the parameters of the 1-D filters have to be chosen so that the performance of the operation is not compromised is investigated. The author presents theoretical results and applies them to some real images to show that the quality of the performance remains the same while the efficiency of the operation improves significantly