Visual Tracking Algorithm

This repository is a implementation of visual tracking algorithm (VTA) proposed in paper Vision Based Robot Assistance in TTTS Fetal Surgery. This algorithm can be used to do segmentation of fetoscopic images for the pourpose of mosaicing.

Data used for testing the algorithm comes form FetReg dataset first introduced in 2021.

Algorithm description

Algorithm can be divided into 5 steps. By following them we should get expected results presented in the paper.

• Extract green channel

  Firstly algorithm extracts green channel from RGB images because this channel has higher contrast between the vascular structures and the background according to this and this papers.

• Apply median blur filter

  Than to remove noise median blur filter is applied.

• Compute CLAHE

  Next, to enhance the contrast between the different structures in the image the algorithm computes a Contrast Limited Adaptive Histogram Equalization (CLAHE).

• Apply thresholding binarization

  After that an adaptive thresholding binarization is applied using the algorithm in here

• Postprocessing filters

  Lastly algorithm applies two postprocessing filters:
  • Specle filter - to eliminate the small componentsof the binary image
  • Morphological closing - filter to fill the small holes in the detected structures

Requirements

• OpenCV