DoGLWDeblending¶
Introduction¶
The DoGLWDeblendingTask
class implements the Laplacian of Gaussian maxim plus Watershed deblending task.
The Laplacian of Gaussian method is used to find the local maxima, and the Watershed method is used to
segment the cluster in the basins identified by the local maxima
Algorithm¶
The algorithm is implemented in the do_guass_laplace_watershed_deblending()
and do_cluster_glw_segmentation()
functions
- The
do_guass_laplace_watershed_deblending()
implements the top level algorithm for the guass_laplace_watershed_deblending: - each parent cluster in the cluster_list is partitioned by the
do_cluster_glw_segmentation()
function. - the parent cluster with his children clusters are used to build
DeblendedProducts
object - a list of class:DeblendedProducts objects is returned
- each parent cluster in the cluster_list is partitioned by the
- The
- The actual parent cluster deblending is implemented in the
do_cluster_glw_segmentation()
function: - The local maxima are obtained by the (
BlobDetectionGaussLaplace
) class - The watershed segmentation on the corresponding local maxima is done using the
watershed()
function
- The local maxima are obtained by the (
- The actual parent cluster deblending is implemented in the
Parameters¶
h_frac
: sets the width of the kernel as h=h_frac*sqrt(r_max^2+r_cluster^2)h_min
: kernel width obtained byh_frac
can not be lower thanh_min
gl_th_rel
: relative threshold for the GaussLaplace local maxima detectionmin_size
: sets the minimum size in pixels to perform a deblending
conf file section¶
[ task: glw_deblending: start]
h_frac = 0.25
h_min = 1.0
gl_th_rel = 0.1
min_size = 9
verbose = False
plot = False
[ task: glw_deblending: stop]