from typing import Literal, Optional
from phenotypic.abc_ import PrefabPipeline
from phenotypic import ImagePipeline
from phenotypic.enhance import CLAHE, GaussianBlur, MedianFilter, ContrastStretching
from phenotypic.detect import OtsuDetector
from phenotypic.grid import GridApply
from phenotypic.refine import (
BorderObjectRemover,
SmallObjectRemover,
LowCircularityRemover,
ReduceMultipleGridObjects,
ResidualOutlierRemover,
)
from phenotypic.correction import GridAligner
from phenotypic.measure import (
MeasureColor,
MeasureShape,
MeasureIntensity,
MeasureTexture,
)
[docs]
class GridSectionPipeline(PrefabPipeline):
"""Detect and measure colonies using per-section processing on grid plates.
Applies a sub-pipeline independently to each grid section, enabling
section-specific thresholds and parameters. Useful when colony properties
vary across the plate (e.g., different strains in different wells).
Steps:
1. GaussianBlur + CLAHE — global preprocessing
2. OtsuDetector — initial global detection
3. BorderObjectRemover, SmallObjectRemover — cleanup
4. GridAligner — straighten the grid
5. GridApply — apply per-section sub-pipeline
6. ReduceMultipleGridObjects — final grid refinement
Measurements: MeasureShape, MeasureColor, MeasureIntensity, MeasureTexture.
Best For:
- Plates where colony properties vary significantly across wells.
- Pre-tiled grid sections that need independent processing.
- Experiments with different strains or conditions in each well.
Consider Also:
- :class:`HeavyOtsuPipeline` when uniform processing across the
plate is sufficient.
- :class:`RoundPeaksPipeline` for a faster approach on consistent
round colonies.
See Also:
:doc:`/explanation/prefab_pipelines_guide` for guidance on choosing
a prefab pipeline.
"""
[docs]
def __init__(
self,
gaussian_sigma: int = 10,
gaussian_mode: str = "reflect",
gaussian_truncate: float = 4.0,
clahe_kernel_size: int | None = None,
clahe_clip_limit: float = 0.01,
median_mode: str = "nearest",
median_cval: float = 0.0,
otsu_ignore_zeros: bool = True,
otsu_ignore_borders: bool = True,
border_remover_size: int | float | None = 50,
circularity_cutoff: float = 0.6,
small_object_min_size: int = 100,
outlier_axis: Optional[int] = None,
outlier_stddev_multiplier: float = 1.5,
outlier_max_coeff_variance: int = 1,
aligner_axis: int = 0,
aligner_mode: str = "edge",
section_blur_sigma: int = 5,
section_blur_mode: str = "reflect",
section_blur_truncate: float = 4.0,
section_median_mode: str = "nearest",
section_median_cval: float = 0.0,
section_contrast_lower_percentile: int = 2,
section_contrast_upper_percentile: int = 98,
section_otsu_ignore_zeros: bool = True,
section_otsu_ignore_borders: bool = True,
grid_apply_reset_enh_matrix: bool = True,
small_object_min_size_2: int = 100,
color_white_chroma_max: float = 4.0,
color_chroma_min: float = 8.0,
color_include_XYZ: bool = False,
texture_scale: int | list[int] = 5,
texture_quant_lvl: Literal[8, 16, 32, 64] = 32,
texture_enhance: bool = False,
texture_warn: bool = False,
benchmark: bool = False,
**kwargs,
):
"""
Initializes the GridSectionPipeline with customizable operations and measurements.
Args:
gaussian_sigma (int): Standard deviation for Gaussian kernel in initial smoothing.
gaussian_mode (str): Mode for handling image boundaries during Gaussian smoothing.
gaussian_truncate (float): Truncate the Gaussian kernel at this many standard deviations.
clahe_kernel_size (int | None): Size of kernel for CLAHE. If None, automatically calculated.
clahe_clip_limit (float): Contrast limit for CLAHE.
median_mode (str): Boundary mode for median filter.
median_cval (float): Constant value for median filter when mode is 'constant'.
otsu_ignore_zeros (bool): Whether to ignore zero pixels in Otsu thresholding.
otsu_ignore_borders (bool): Whether to ignore border objects in Otsu detection.
border_remover_size (int | float | None): Size of border region where objects are removed.
circularity_cutoff (float): Minimum circularity threshold for objects to be retained.
small_object_min_size (int): Minimum size of objects to retain in first removal step.
outlier_axis (Optional[int]): Axis for outlier analysis. None for both, 0 for rows, 1 for columns.
outlier_stddev_multiplier (float): Multiplier for standard deviation in outlier detection.
outlier_max_coeff_variance (int): Maximum coefficient of variance for outlier analysis.
aligner_axis (int): Axis for grid alignment (0 for rows, 1 for columns).
aligner_mode (str): Mode for grid alignment rotation.
section_blur_sigma (int): Standard deviation for Gaussian kernel in section-level detection.
section_blur_mode (str): Mode for Gaussian smoothing in section-level detection.
section_blur_truncate (float): Truncate for Gaussian kernel in section-level detection.
section_median_mode (str): Boundary mode for median filter in section-level detection.
section_median_cval (float): Constant value for median filter in section-level detection.
section_contrast_lower_percentile (int): Lower percentile for contrast stretching in sections.
section_contrast_upper_percentile (int): Upper percentile for contrast stretching in sections.
section_otsu_ignore_zeros (bool): Whether to ignore zeros in section-level Otsu detection.
section_otsu_ignore_borders (bool): Whether to ignore borders in section-level Otsu detection.
grid_apply_reset_enh_matrix (bool): Whether to reset detect_mat before applying section-level pipeline.
small_object_min_size_2 (int): Minimum size of objects to retain in second removal step.
color_white_chroma_max (float): Maximum white chroma value for color measurement.
color_chroma_min (float): Minimum chroma value for color measurement.
color_include_XYZ (bool): Whether to include XYZ color space measurements.
texture_scale (int | list[int]): Scale parameter(s) for Haralick texture features.
texture_quant_lvl (Literal[8, 16, 32, 64]): Quantization level for texture computation.
texture_enhance (bool): Whether to enhance image before texture measurement.
texture_warn (bool): Whether to warn on texture computation errors.
benchmark (bool): Indicates whether benchmarking is enabled across the pipeline.
"""
ops = {
"blur" : GaussianBlur(
sigma=gaussian_sigma, mode=gaussian_mode, truncate=gaussian_truncate
),
"clahe" : CLAHE(kernel_size=clahe_kernel_size,
clip_limit=clahe_clip_limit),
"median filter" : MedianFilter(mode=median_mode,
cval=median_cval),
"detection" : OtsuDetector(
ignore_zeros=otsu_ignore_zeros, ignore_borders=otsu_ignore_borders
),
"border_removal" : BorderObjectRemover(
border_size=border_remover_size),
"low circularity remover" : LowCircularityRemover(
cutoff=circularity_cutoff),
"small object remover" : SmallObjectRemover(
min_size=small_object_min_size),
"Reduce by section residual error": ReduceMultipleGridObjects(),
"outlier removal" : ResidualOutlierRemover(
axis=outlier_axis,
stddev_multiplier=outlier_stddev_multiplier,
max_coeff_variance=outlier_max_coeff_variance,
),
"align" : GridAligner(axis=aligner_axis,
mode=aligner_mode),
"section-level detect" : GridApply(
ImagePipeline(
{
"blur" : GaussianBlur(
sigma=section_blur_sigma,
mode=section_blur_mode,
truncate=section_blur_truncate,
),
"median filter" : MedianFilter(
mode=section_median_mode,
cval=section_median_cval
),
"contrast stretching": ContrastStretching(
lower_percentile=section_contrast_lower_percentile,
upper_percentile=section_contrast_upper_percentile,
),
"detection" : OtsuDetector(
ignore_zeros=section_otsu_ignore_zeros,
ignore_borders=section_otsu_ignore_borders,
),
}
),
reset_enh_matrix=grid_apply_reset_enh_matrix,
),
"small object remover 2" : SmallObjectRemover(
min_size=small_object_min_size_2
),
"grid_reduction" : ReduceMultipleGridObjects(),
}
meas = {
"MeasureColor" : MeasureColor(
white_chroma_max=color_white_chroma_max,
chroma_min=color_chroma_min,
include_XYZ=color_include_XYZ,
),
"MeasureShape" : MeasureShape(),
"MeasureIntensity": MeasureIntensity(),
"MeasureTexture" : MeasureTexture(
scale=texture_scale,
quant_lvl=texture_quant_lvl,
enhance=texture_enhance,
warn=texture_warn,
),
}
super().__init__(ops=ops, meas=meas, benchmark=benchmark, **kwargs)
