Adaptive Thresholding Segmentation Algorithm

Description

I employed adaptive thresholding to correct for uneven illumination in the image. Because intensity of different structures in the cell varied significantly, I used a large window size to ensure that the same structure in different areas was thresholded the same way.

To ensure smooth differences in thresholding and to avoid discontinuities in thresholding due to different thresholds in different local windows, I thresholded a large window (linear dimensions equal to 1/10th the image’s linear dimensions), but I only used the result in a smaller sub-window (1/40th the images linear dimension).

I tried different values for this parameter, but the result was reasonably robust to the choice of parameter, as long as the sizes of the large window and small window were sufficiently different.

MATLAB Implementation of segmentation-based disparity averaging

Efficient Graph-Based Image Segmentation