Image Fusion Based Local Graph Matching For Plant Cell Tracking

Developing robust tracking algorithms for four-dimensional (3D + time) plant cells and their division is extremely important for obtaining spatiotemporal measurements of cell behavior patterns. The tracking of plant cells across noisy microscopy image sequences is very challenging, because plant cells in noisy region cannot be correctly segmented and cause serious errors in subsequent cell tracking procedure. This paper proposed an image fusion based local graph matching method to track plant cells. First, the nonsubsampled contourlet transform sparse representation (NSCTSR) image fusion method is used to fusing a confocal plant cell image stack into a single image, which has the higher contrast of image quality and richer information than any single image in the image stack. Second, the local graph matching approach is used to track the plant cells in the fused images, by exploiting the cells' local graph structure and contextual information. The experimental results demonstrate that the proposed method can improve the quality of the cell image, and reach a higher tracking accuracy for plant cells than the previous method.