Fig. 2

LSM for tissue contour. a–c Section of a developing flower at the boundary between the peripheral zone and the forming sepal. a Tissue contour detected with the “Edge Detect” process of the MorphoGraphX software. The contour doesn’t enter in deep creases (asterisk) and presents crests (instead of valleys) above the anticlinal walls (arrows). b Tissue contour detected with the level set method. The green contour is an initialisation, while the red contour is the result of the level set contour detection (\(\alpha =1\), \(\beta =0\)). c The effect of the accelerating term weighted by the parameter \(\alpha\) is to “push” the contour into profound valleys on the surface: green contour \(\alpha =0\), red contour \(\alpha =1\) and all other parameters are the same and default values. d, e The effect of the smoothing term weighted by the parameter \(\beta\): d \(\beta =0\) and e \(\beta =1\) and all other parameters are the same and default values. f, g Accurate surface detection allows accurate signal projection on the surface. Here is an example of membrane signal (f) and microtubule signal (g) projected using MorphoGraphX on a surface detected with the level set method based on the membrane signal. h–j The surface detected with the level set method allows to do accurate curvature maps, which can be segmented in order to determine organ boundaries. Here we present the curvature map and its segmentation done with MorphoGraphX: h Average curvature map (on a neighbourhood of 10 μm). i Manual seeding of the curvature map. j Watershed segmentation of the map