Hierarchical optimum edge detection for linear array sensors

In the traditional approach to optimum edge detection, a single filter is defined based on cumulatively, but not individually optimizing all the desirable (e.g. detection, localization) criteria. We pursue the application of both detection-optimum and localization-optimum filters, and utilize the results hierarchically. The method involves finding the range of pixels where the edge transition occurs using the detection-optimum result, and then within this range, to use the localization-optimum result to identify the location. The extent of the range is determined by where the SNR for the localization-optimum filter exceeds the detection-optimum filter, hence the overall result is the best combination of performance where each criterion is optimally satisfied. This approach achieves a modest improvement over the application of a single detection-localization optimum filter, at the expense of doubling the required computational load.