Computer Vision News - March 2021

16 Computer Vision Tool But what if we wanted to try segmenting only the cells’ boundaries? The figure below shows examples of modifying the threshold using the dedicated window. This example can be further improved by setting the correct scale for the analyzed image. On microscopy images, it can be useful to use the scale bar at the bottom, if present. You just need to go on the line icon and draw a line of the same length of the bar and then press on Analyze -> Set scale and write the corresponding information on the Known Distance (100) and Unit of length (um) field. Another useful technique offered by ImageJ is thresholding or segmentation of areas. If we take as sample an RGB image, before applying a threshold, we first need to split it into the different channels - to do this in ImageJ you just need to go on Image -> Color -> Split channels. Now three new images should appear on the screen, and we can select one (in the example below, the blue channel), then press on Adjust -> Threshold to visualize a window with the histogram and sliders that allow to modify the threshold and observe the obtained results directly on the desired image. Note that the default values for the threshold are set by ImageJ based on the histogram, so the first attempt should be close to a plausible result. The image below features the blue channel segmented using the default threshold and it does a pretty good job at isolating the cells.