Automatic Cell Detection and Segmentation.

The package contains the following files and folders:

• ./batch_watershed_single.m: The batch program to detect and segment single cell automatically.
• ./batch_watershed_multiple.m: The batch program to detect and segment multiple cells automatically.
• ./saliency_cvpr09.m: Detects salient objects proposed by IG method [1].
• ./saliency_cvpr09_S.m: Detects salient objects by Saturation.
• ./rgb2hsi.m: Converts an RGB image to HSI.
• ./expectation_variance.m: Computes the expectation and variance of a gray-level image.
• ./remove_corner_noise.m: Removes the possible exist corner noise caused by illumination.
• ./extract_single_cell_region.m: Extracts the biggest region for single cell.
• ./size_objects.m: Computes the sizes of all objects.
• ./size_otsu.m: Uses OTSU's thresholding method to classify the sizes of the objects.
• ./modify_binary_image.m: Removes the possible exist small noise according to the size.
• ./watershed_single.m: Marker-controlled watershed algorithm for single cell segmentation.
• ./watershed_multiple.m: Marker-controlled watershed algorithm for multiple cells segmentation.
• ./DATA: This folder contains the benchmark DATASET for automatic cells detection and segmentation with MicroscopicImages and GroundTruth containing single and multiple resepectively.
• ./RESULTS: This folder contains the RESULTS for automatic cells detection and segmentation.
• ./tools: This folder contains some useful scripts.

1. Put the microscopic images with a single cell in ./DATA/MicroscopicImages/single/ folder;
2. Run ./batch_watershed_single.m to detect and segment the cell into ./RESULTS/single/ folder with the running time records in ./RESULTS/single.time.

1. Put the microscopic images with multiple cells in ./DATA/MicroscopicImages/multiple/ folder;
2. Run ./batch_watershed_multiple.m to detect and segment the cells into ./RESULTS/multiple/ folder with the running time records in ./RESULTS/multiple.time.

[1] Radhakrishna Achanta, Sheila Hemami, Francisco Estrada and Sabine Susstrunk. Frequency-tuned Salient Region Detection. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2009.

This code is free for use in research projects. If you publish results obtained using this code, please consider citing relevant paper:

Haiyong Zheng, Nan Wang, Zhibin Yu, Zhaorui Gu, Bing Zheng. Robust and automatic cell detection and segmentation from microscopic images of non-setae phytoplankton species. IET Image Processing, 2017, DOI: 10.1049/iet-ipr.2017.0127.

If you have any question, please feel free to contact Haiyong Zheng ([email protected]).