An initial step: Similar to a single-resolution edge detection method, the first step of this method starts
with Sobel edge detection and followed by the Canny edge detection. However, this time a large
value of smoothing parameter of the Canny edge detection is used (i.e.
√
10). A large value
of smoothing parameter is used because in an initial step we only need to roughly estimate the
boundary of the algae in an image. The true boundary will be detected in the next refinement step.
A refinement step: After separating the algae body from a background, only the background of the
image will be heavily smoothed by using a Gaussian lowpass filter of size 20 × 20 pixels with
sigma equal to 0.5 in order to suppress all unwanted objects and illumination artefacts in the
image background. The foreground of the image (i.e. the algae body) is left unsmoothed because
we want to preserve as much as possible its edge details.
After a smoothing process, the smoothed image will simply be segmented by the single-resolution
edge detection method described above. A multi-resolution edge detection method derives its
name from the fact that edge detection is performed on an image with different smoothing reso-
lutions of foreground and background regions.