Supplementary Figure 1: Segmentation of fluorescent bacteria.

a: input image, illustrating fluorescence heterogeneity within and between cells. b: Image of edges computed with a sigma transform of the gaussian smoothed image. Along the cell contours, fluorescence intensities are maximal at the interface between cells and background, and minimal at the boundary between cells. c: regions obtained by a watershed algorithm computed on image b. For each region, the median value of the fluorescence intensity within this region in image A is displayed through a color code (see color bar below). d: Interfaces between the different regions shown in C. For each interface, the color indicates the value of the first decile of the distribution of pixel intensity in image B at this interface. e: Maximum Eigenvalue of the Hessian transform (MEH). MEH is maximal at the interfaces between cells. f: Regions obtained by watershed algorithm computed on E within the foreground region defined using image c. For each region, the color indicates the median intensity of image A within this region. g: mean intensity of image E divided by mean intensity of image A at the interfaces between foreground regions displayed in F. H Final result of the segmentation process.