Cell Tracking: The Study On Gray Level Cell Image For Improving The Accuracy Of Cell Segmentation

The cell tracking is one of the important research area in cell study, biology andmedicine. It is the new research challenge arising in digital image processing andcomputer vision field in recent decade. Several existing cell tracking system assumehigh accuracy of cell segmentation, while if there are errors in cell segmentationmodule, there would be severe negative effect on the accuracy and efficiency of celltracking results. When we face the truth that the accuracy of cell tracking can’t fulfillthe designated standard, in this thesis we focus on improving the accuracy of cellsegmentation, following the analysis and comparison among the main available cellsegmentation and tracking algorithm. To solve this problem, we analyzed thedifference between cell tracking and general object tracking platforms, as well as themain reasons that causes the difficulty for cell segmentation accurately.Furthermore, we provide the literature review on the main available cell trackingalgorithm in this thesis, the comparison between specific cell tracking and generalobject tracking. Then we introduce and analyze the state space graph andViterbi-based cell tracking system in detail.For the purpose to improve the accuracy of cell segmentation, we proposed a novelapplication of shifted bi-Gaussian image gray level histogram matching method on thestudy of cell segmentation. Also we use mathematical morphological operation toassist the cell segmentation work for better cell segmentation results.The validity of our proposed method is tested by using shifted bi-Gaussian imagegray level histogram matching and mathematical morphological operation forimproving the accuracy of cell segmentation using Viterbi-based cell tracking system.We test two time-lapse cell image sequences, N2DH-GOWT1sub-database,182frames totally, downloaded from International Symposium on BIOMEDICALIMAGING2013(ISBI2013). We also introduce our proposed testing system. Ourproposed method in this paper improved the accuracy of cell segmentation, validatedby the comparison of the cell segmentation results by these two time-lapse cellsequences.