Depth estimation in monocular Breast Self-Examination image sequence using optical flow

In this paper, we study the depth estimation for image sequence with small displacements as in Breast Self Examination (BSE). We utilized its Lucas-Kanade optical flow vectors, the concept of divergence and focus of expansion to estimate the apparent depth level for each frame. Moreover, orientation binning is also introduced to supplement its invariance to translation. The experiment used an actual BSE performance and the results show its effectiveness in predicting palpation depth level. This algorithm has shown to be in realtime implementation with a frame rate of 30 frames per second that is very useful for implementing the computer vision-based BSE guidance system.