Automatic system for stitching microscopic images using OpenPano

Analyzing pap smear images by manual reading through a microscope is a challenge for doctors especially in distinguishing epithelial cells from inflammatory cells. A computer-based application is required for cervical cancer screening and displaying the results of reading automatically pap smear images using a microscope. This study aims to develop a module for stitching microscopic images including pap smear images which are collected from a microscope. A module for automatically stitching pap smear images is an important component in pap smear images management system. There are three tasks in stitching several pap smear images which were implemented in this study. Those three tasks include 1) feature extraction with scale-invariant feature transform, 2) estimation of homography or affine transformation by random sample consensus and 3) optimization of connection between images with bundle adjustment and blending. We use OpenPano to implement those three tasks. The system has successfully stitched the pap smear images with the feature overlap among the images greater than 10%. The stitched pap smear image has good quality after smoothing using the vertical blending step. In addition, blurry pixels on in some spots disappeared after performing the stitching error. The image quality has been improved after the multiband blending using 16-band and 128-band which shows the significant improvement on the 128-band. The system is expected to assist doctors in pap smear image identification.