This catalog covers every public module declared in src/ops/mod.rs, together with image-processing capabilities exposed elsewhere in the crate. It groups related operations by purpose; it does not treat every helper function as a separate algorithm.
Most image kernels use rank-three z/y/x buffers. A singleton z dimension represents 2D data where the chosen algorithm supports it. Dtype support is operation-specific: the default BlockOp contract accepts f64, while many concrete ops explicitly accept other types. Never infer type support from the catalog’s topic alone.
Execution labels. A planned route declares reads and writes to Blockflow; a resident helper expects the caller’s in-memory arrays. A planned whole-axis or global solve can still have a large memory requirement. The topic pages call out these distinctions rather than treating every exported API as equally scalable.
Choose by dependency shape
A halo kind is a statement about the computation, not a speed setting. Start with the algorithm whose result you need, then choose a block geometry that can satisfy it. The halo design page derives these forms; the topic pages explain their operation-specific parameters and limits.
| No spatial neighborhood |
Pointwise maps and tuples, calibrated camera correction and NNLS |
Every output uses co-located values; additional channels still cost buffers and explicit source dependencies. |
| Finite symmetric or asymmetric stencil |
Gaussian, direct convolution and rank, primitive morphology |
Kernel support and origin specify lower/upper sides. Even windows and off-center kernels are not necessarily symmetric. |
| Block-dependent grant |
Sampled local statistics |
Globally fixed sample centers meet each core at different distances. A per-block halo can avoid the worst-case grant while preserving the same sample lattice. |
| Tile-aligned stencil |
FFT convolution and transform banks |
Anchored transform tiles need extra source coverage when compute block edges do not align with tile edges. |
| Mapped source region |
Resampling, warping and plane sampling, split lattice statistics |
Output coordinates and source coordinates differ; dependencies are bounding source regions rather than one radius around equal-index cores. |
| Complete axis or plane |
Exact Euclidean distance, spline coefficients, DIC integration/destriping |
A line optimum, line solve or plane Fourier inverse can depend on samples arbitrarily far along those axes. |
| Local evidence with global decisions |
Components/hysteresis, global thresholds, graph solves |
Small local facts can be merged globally; requesting a huge image halo would confuse synchronization with image reads. The reduced state may still be large. |
| Neighboring fragment support |
Voxel/outline painting, capped point distances |
An object anchored outside a core can contribute inside it. Fetch relevant producer fragments, not duplicate object fitting or unrelated image pixels. |
| Iterative local update |
Diffusion, reconstruction, finite-PSF deconvolution |
Each update exchanges current neighboring state. A short one-step halo does not imply a finite global propagation distance. |
Some operations deliberately group several local updates into one application. For example, directional thinning and fixed-pass configuration rewriting declare the transitive reach of the grouped updates. Other workflows insert state boundaries between updates. Both can be correct, but their halo work and stored intermediates differ.
Complete module index
Each row below corresponds to a public ops module. Internal implementation files are covered through their exported parent module. For example, calibrated tensor eigensystems and support-weighted tensors are exported by structure_tensor.
ops::activity |
Pointwise maps and channel arithmetic |
ops::adjacency |
Measurements and relationships |
ops::agglomerate |
Graph and iterative segmentation |
ops::align |
Registration and alignment |
ops::background |
Intensity, background and thresholds |
ops::boundaries |
Measurements and relationships |
ops::camera |
Calibration, unmixing and deconvolution |
ops::classify |
Classification, learned models and evaluation |
ops::cleanup |
Components, watershed and region cleanup |
ops::components |
Components, watershed and region cleanup |
ops::configuration |
Morphology, topology and distance |
ops::convolve |
Filters, derivatives and feature images |
ops::coordinates |
Coordinates, painting and row operations |
ops::cost |
Specialised imaging and tracing |
ops::deconvolution |
Calibration, unmixing and deconvolution |
ops::deconvolve |
Calibration, unmixing and deconvolution |
ops::destripe |
Specialised imaging and tracing |
ops::detect |
Components, watershed and region cleanup |
ops::dic |
Specialised imaging and tracing |
ops::diffuse |
Filters, derivatives and feature images |
ops::directional |
Morphology, topology and distance |
ops::distance |
Morphology, topology and distance |
ops::edges |
Filters, derivatives and feature images |
ops::eggshell |
Specialised imaging and tracing |
ops::element |
Morphology, topology and distance |
ops::envelope |
Morphology, topology and distance |
ops::exposure |
Intensity, background and thresholds |
ops::features |
Filters, derivatives and feature images |
ops::fft |
Resampling, warps and sampling |
ops::fill |
Morphology, topology and distance |
ops::intensity_statistics |
Intensity, background and thresholds |
ops::interest |
Spots, templates and fitted objects |
ops::interior |
Spots, templates and fitted objects |
ops::interior_partition |
Spots, templates and fitted objects |
ops::label |
Components, watershed and region cleanup |
ops::label_energy |
Graph and iterative segmentation |
ops::lattice |
Intensity, background and thresholds |
ops::lattice_samples |
Intensity, background and thresholds |
ops::level_set |
Graph and iterative segmentation |
ops::local |
Intensity, background and thresholds |
ops::measure |
Measurements and relationships |
ops::mixing |
Pointwise maps and channel arithmetic |
ops::morphology |
Morphology, topology and distance |
ops::normalise |
Intensity, background and thresholds |
ops::path_opening |
Morphology, topology and distance |
ops::plane_sample |
Resampling, warps and sampling |
ops::point_distance |
Measurements and relationships |
ops::psf_fit |
Calibration, unmixing and deconvolution |
ops::rag |
Graph and iterative segmentation |
ops::rag_phases |
Graph and iterative segmentation |
ops::random_walker |
Graph and iterative segmentation |
ops::rank |
Filters, derivatives and feature images |
ops::rasterise |
Coordinates, painting and row operations |
ops::reconstruct |
Morphology, topology and distance |
ops::regional |
Components, watershed and region cleanup |
ops::resample |
Resampling, warps and sampling |
ops::ridge |
Filters, derivatives and feature images |
ops::rows |
Coordinates, painting and row operations |
ops::scikitimage_watershed |
Components, watershed and region cleanup |
ops::segment_split |
Graph and iterative segmentation |
ops::skeleton |
Morphology, topology and distance |
ops::slice_normalise |
Intensity, background and thresholds |
ops::sliding |
Filters, derivatives and feature images |
ops::smooth |
Filters, derivatives and feature images |
ops::spline |
Resampling, warps and sampling |
ops::stack_align |
Registration and alignment |
ops::star_convex |
Spots, templates and fitted objects |
ops::structure_tensor |
Filters, derivatives and feature images |
ops::suppress |
Spots, templates and fitted objects |
ops::surfaces |
Graph and iterative segmentation |
ops::tabulate |
Measurements and relationships |
ops::template_match |
Spots, templates and fitted objects |
ops::threshold |
Intensity, background and thresholds |
ops::tiled_watershed |
Components, watershed and region cleanup |
ops::transform |
Resampling, warps and sampling |
ops::transform_bank |
Filters, derivatives and feature images |
ops::unmixing |
Calibration, unmixing and deconvolution |
ops::vessel_trace |
Specialised imaging and tracing |
ops::voxelize |
Coordinates, painting and row operations |
ops::voxelwise |
Pointwise maps and channel arithmetic |
ops::walk |
Coordinates, painting and row operations |
ops::watershed |
Components, watershed and region cleanup |
Other public image-analysis modules
agreement |
Classification, learned models and evaluation |
classical_centers |
Classification, learned models and evaluation |
feature_index |
Coordinates, painting and row operations |
| forest |
Classification, learned models and evaluation |
gpu |
Classification, learned models and evaluation |
intensity_pyramid |
Resampling, warps and sampling |
label_pyramid |
Resampling, warps and sampling |
model_registry |
Classification, learned models and evaluation |
model_segment |
Classification, learned models and evaluation |
model_training |
Classification, learned models and evaluation |
model_worker |
Classification, learned models and evaluation |
| object_table |
Coordinates, painting and row operations |
points |
Coordinates, painting and row operations |
pyramid |
Resampling, warps and sampling |
sam |
Classification, learned models and evaluation |
sam_training |
Classification, learned models and evaluation |
sampling |
Resampling, warps and sampling |
segmentation_metrics |
Classification, learned models and evaluation |
synthetic |
Classification, learned models and evaluation |
synthetic_tissue |
Classification, learned models and evaluation |
table |
Coordinates, painting and row operations |
track |
Classification, learned models and evaluation |
yolo |
Classification, learned models and evaluation |
yolo3d |
Classification, learned models and evaluation |
Source and API detail
The source module declarations and feature definitions provide exact signatures and compile-time availability. The accompanying inventory.json maps declarations to pages so future additions can be checked for documentation coverage. Coverage here means a discoverable, described module; it is not a claim that every specialized option already has its own tutorial.