An image segmentation algorithm based on the watershed paradigm
Project description
version 1.1.1 presents cleaned-up documentation. The implementation code remains unchanged.
Version 1.1 fixes a bug in the dilate() and erode() methods of the module that caused these methods to misbehave for non-square images. Version 1.1 also includes improvements in the explanatory comments included in the scripts in the Examples directory.
This module is a Python implementation of the Watershed algorithm for image segmentation. The goal of this module is not to compete with the popular OpenCV implementation of the watershed algorithm. On the other hand, the goal here is to provide an alternative framework that is more amenable to experimentation with the logic of watershed segmentation.
Typical usage syntax:
from Watershed import * shed = Watershed( data_image = "orchid0001.jpg", binary_or_gray_or_color = "color", size_for_calculations = 128, sigma = 1, gradient_threshold_as_fraction = 0.1, level_decimation_factor = 16, ) shed.extract_data_pixels() shed.display_data_image() shed.mark_image_regions_for_gradient_mods() #(A) shed.compute_gradient_image() shed.modify_gradients_with_marker_minima() #(B) shed.compute_Z_level_sets_for_gradient_image() shed.propagate_influence_zones_from_bottom_to_top_of_Z_levels() shed.display_watershed() shed.display_watershed_in_color() shed.extract_watershed_contours() shed.display_watershed_contours_in_color() The statements in lines (A) and (B) are needed only for marker-assisted segmentation with the module. For a fully automated implemented of the BLM algorithm, you would need to delete those two statements.