hnccorr.seeds module¶
Seed related components of HNCcorr.
-
class
hnccorr.seeds.LocalCorrelationSeeder(neighborhood_size, keep_fraction, padding, grid_size)[source]¶ Bases:
objectProvide seeds based on the correlation of pixels to their local neighborhood.
Seed pixels are selected based on the average correlation of the pixel to its local neighborhood.For each block of grid_size by grid_size pixels, the pixel with the highest average local correlation is selected. The remaining pixels in each block are discarded. From the remaining pixels, a fraction of _seed_fraction pixels, those with the highest average local correlation, are kept and attempted for segmentation.
The local neighborhood of each pixel consist of the pixels in a square of width _neighborhood_size centered on the pixels. Pixel coordinates outside the boundary of the movie are ignored.
- Variables
_current_index (int) – Index of next seed in _seeds to return.
_excluded_pixels (set) – Set of pixel coordinates to excluded as future seeds.
_grid_size (int) – Number of pixels per dimension in a block.
_keep_fraction (float) – Percentage of candidate seed pixels to attempt for segmentation. All other candidate seed pixels are discarded.
_movie (Movie) – Movie to segment.
_neighborhood_size (int) – Width in pixels of the local neighborhood of a pixel.
_padding (int) – L-infinity distance for determining which pixels should be padded to the exclusion set in exclude_pixels().
_seeds (list[tuple]) – List of candidate seed coordinates to return.
-
__init__(neighborhood_size, keep_fraction, padding, grid_size)[source]¶ Initializes a LocalCorrelationSeeder object.
-
_compute_average_local_correlation(pixel, valid_neighbors)[source]¶ Compute average correlation between pixel and neighbors.
-
_select_best_per_grid_block(scores)[source]¶ Selects pixel with highest score in a block of grid_size pixels per dim.
-
exclude_pixels(pixels)[source]¶ Excludes pixels from being returned by next() method.
All pixels within in the set pixels as well as pixels that are within an L- infinity distance of _padding from any excluded pixel are excluded as seeds.
Method enables exclusion of pixels in previously segmented cells from serving as new seeds. This may help to prevent repeated segmentation of the cell.
- Parameters
pixels (set) – Set of pixel coordinates to exclude.
- Returns
None
-
next()[source]¶ Provides next seed pixel for segmentation.
Returns the movie coordinates of the next available seed pixel for segmentation. Seed pixels that have previously been excluded will be ignored. Returns None when all seeds are exhausted.
- Returns
Coordinates of next seed pixel. None if no seeds remaining.
- Return type
tuple or None
-
class
hnccorr.seeds.NegativeSeedSelector(radius, count)[source]¶ Bases:
objectSelects negative seed pixels uniformly from a circle around center seed pixel.
Selects _count pixels from a circle centered on the center seed pixel with radius _radius. The selected pixels are spread uniformly over the circle. Non-integer pixel indices are rounded to the closest (integer) pixel. Currently only 2-dimensional movies are supported.
- Variables
_radius (float) – L2 distance to center seed.
_count (int) – Number of negative seed pixels to select.
-
class
hnccorr.seeds.PositiveSeedSelector(max_distance)[source]¶ Bases:
objectSelects positive seed pixels in a square centered on center_seed.
Selects all pixels in a square centered on center_seed as positive seeds. A pixel is selected if it is within a Chebyshev distance (L-Inf) of _max_distance from the center seed pixel.
- Variables
_max_distance (int) – Maximum L-Inf distance allowed.