TV-L1 based fast and robust mosaicing of cystoscopic images
Résumé
Image mosaicing helps in diagnosis and follow-up of lesions in cystoscopic bladder image sequences by increasing the field of view. Optical flow methods have been used for various scenes, whereas they were barely experimented for data with poor texture and blurry images like that of endoscopic data. In this paper, we present the use of total variational optical flow method (TV-L1) for such high texture variability and blur. This method is compared to other optical flow methods on phantom images simulating realistic patient data. This comparison and the result on patient data show the advantage of the TV-L1 method over the other methods in terms of registration accuracy, robustness and time.