Colposcopic image registration using opponentSIFT descriptor

This work presents a colposcopic image registration system able to help physicians for cervical cancer diagnosis. The goal is to make registration between the cervical tissue throughout the whole temporal sequence. Recent digital images processing works, suggested using feature points to compute the tissue displacement. These methods achieve good results, because they are fast and do not need any segmentation, but to date, all methods based on feature points are sensitive to light change and reflections which are frequently current in colposcopic images. To solve this problem, we propose to apply the opponentSIFT descriptor which describes features point in the opponent color space. Experimental results show the robustness of this descriptor in colposcopic images registration in comparison with other descriptors.