Detecting Edges in Noisy Face Database Images

In this paper, a morphological-based system for detecting edges in reallife images is presented. The corner stone for this system is the hit-miss transform, which provides good performance in reallife images under noise conditions. The classical implementation of this transform suffers from drawbacks that are tackled in this paper. The new modified hit-miss transform is introduced to provide better edge detection and noise handling. The performance of the modified hit-miss transform is evaluated by comparing it with classical edge detection techniques. This system is implemented as a pre-processing stage for detecting faces belonging to the Manchester face database in order to remove noise and background regions from them.