In contactless metrology sub-pixel edge detectors allow for high precision dimensional measurements of lengths and positions of objects. Several classes of detectors exist with various assumptions about the edge model and different requirements on computational resources. Moment-based detectors are fast to compute and offer potential for parallelisation. However, being based on a step-edge model, moment based detectors often introduce systematic errors with blur from diffraction or defocus. In this paper, the performance of the 1D spatial moments detector is analysed under influence of Gaussian blur and an error correction term is proposed, reducing measurement errors for blurred images by factors of 2 - 7.
展开▼
