Creating a virtual slide map of sputum smears by auto-stitching

Under a microscope, only a small field of a slide can be seen through the eye piece. A digital replication of a large section of the smear on a slide is desirable for several applications, including image-based auto-positioning for automated microscopy and telepathology. To achieve this, images of the various fields on the slide can be captured and stitched automatically to form a virtual slide map. We compared two auto-stitching methods, namely the geometric hashing and the scale invariant feature transform (SIFT), applied to images of tuberculosis-infected sputum smears. Experimental results show good agreement between the two methods.