SAIRPICO Deconvolution¶

sairpico_deconvolution shows how BioImageFlow wraps SAIRPICO command-line tools in a microscopy restoration workflow. It starts from a supplied microscopy crop, generates a point-spread function, runs denoising, feeds the generated PSF into Richardson-Lucy deconvolution, and records image-quality metrics.

Use this workflow when you want a reproducible deconvolution pipeline whose external command calls are visible in the graph. Install the real SAIRPICO binaries before executing the command; the workflow graph keeps every generated image and command output explicit.

Microscopy crop and generated Gaussian PSF for SAIRPICO deconvolution — The generated PSF is an explicit workflow artifact, not an implicit parameter hidden inside the deconvolution call.¶

Run the workflow with a supplied crop:

python example_workflows/sairpico_deconvolution/workflow.py --input-image data/13432_fish_crop.tif

FISH CIL crops are convenient lightweight inputs for this tutorial. For a production deconvolution run, use a crop and PSF model that match the microscope and acquisition settings.

Pipeline walkthrough¶

flowchart LR
  image[Microscopy crop]:::source --> denoise[SAIRPICO median denoising]:::process
  image --> psf[Generate PSF]:::process
  denoise --> join[Pair denoised image with PSF]:::process
  psf --> join
  join --> rl[Richardson-Lucy deconvolution]:::process
  image --> metrics[Sharpness and residual-noise metrics]:::metric
  denoise --> metrics
  rl --> metrics
  classDef source fill:#e7f0ff,stroke:#4b73b9,color:#1b2f55
  classDef process fill:#edf8ef,stroke:#4d8f5b,color:#173d20
  classDef metric fill:#ffeceb,stroke:#b85b52,color:#4d201c

The explicit merge node matters because deconvolution consumes two upstream products: the denoised image and the generated PSF. That makes the dependency visible in exported workflow graphs and in cached results.

What you will inspect¶

Documentation preview of denoised and deconvolved microscopy images — The documentation preview illustrates the pair of image artifacts to inspect after execution: the SAIRPICO denoised image and the Richardson-Lucy deconvolution result.¶

The metrics table is a compact numerical summary, but the images are the primary review artifact. If deconvolution creates ringing, amplifies background structure, or sharpens noise, change the PSF and iteration settings before trusting the output.