---
vault_clearance: EUCLID
halo:
  classification: INTERNAL
  confidence: HIGH
  front: "22_Project_StarSight"
  custodian: "The Architect"
  created: 2026-03-30
  updated: 2026-04-01
  containment: "BOOK — bibliography + software + benchmarks for bioimaging vision"
---

# BOOK — StarSight (Bibliography Of Online Knowledge)

Canonical **bibliography, software registry, and benchmark index** for **22_Project_StarSight**. Convention: [BOOK_Protocol.md](../BOOK_Protocol.md).

**Operational map (pipelines, cultivator, proof ladder):** [FORM.md](FORM.md). **Triad:** [README.md](README.md) · [BOUNTY_BOARD.md](BOUNTY_BOARD.md) · [WORLDLINE.md](WORLDLINE.md).

**RESTRICTED exemplar (read paths, do not mirror to edge):** Wet-lab ingest under **10** — `10_Project_DiscordIntoSymphony/data/usb_ingest_STORE_N_GO_2026-03-30/` — [BOOK_Protocol.md](../BOOK_Protocol.md) wet-lab subsection; [HALO_PROTOCOL.md](../HALO_PROTOCOL.md).

Rows below use **verified DOIs** where cited; software without a single canonical paper lists **official docs** and a **Curate:** note for a stable citation.


### Local registry slice (EYE / STAFF / STARS)

| Surface | Pointers |
|---------|----------|
| **EYEs** | Runs: [`README.md`](README.md) / [`WORLDLINE.md`](WORLDLINE.md) (if present). Registry: [`EYE_PROTOCOL.md`](../EYE_PROTOCOL.md) |
| **STAFF** | Runnable tools: [`STAFF_catalogue.json`](../STAFF_catalogue.json) — filter `project_dir` for this folder. |
| **STARS** | This file; rules: [`BOOK_Protocol.md`](../BOOK_Protocol.md). |
| **Audit sheet** | [`LOGGING_AND_REGISTRY_CHECKLIST.md`](../99_Archive/root_reports/2026-04/LOGGING_AND_REGISTRY_CHECKLIST.md) |


---

## 1. Primers, methods reviews, and landmark segmentation papers

| ID | Kind | Title / note | Identifier |
|----|------|--------------|------------|
| SS-B1 | Software | QuPath: open source software for digital pathology | [10.1038/s41598-017-17204-5](https://doi.org/10.1038/s41598-017-17204-5) |
| SS-B2 | Software | Cellpose: a generalist algorithm for cellular segmentation | [10.1038/s41592-020-01001-x](https://doi.org/10.1038/s41592-020-01001-x) |
| SS-B3 | Software | CellProfiler: image analysis software for identifying and quantifying cell phenotypes | [10.1186/gb-2006-7-10-r100](https://doi.org/10.1186/gb-2006-7-10-r100) |
| SS-B4 | Software | Ilastik: interactive learning and segmentation toolkit | [10.1038/nmeth.2082](https://doi.org/10.1038/nmeth.2082) |
| SS-B5 | Software | napari: multi-dimensional image viewer for Python | [10.1371/journal.pcbi.1008980](https://doi.org/10.1371/journal.pcbi.1008980) |
| SS-B6 | Methods | scikit-image: image processing in Python | [10.7717/peerj.453](https://doi.org/10.7717/peerj.453) |
| SS-B7 | Segmentation | Cell detection with star-convex polygons (StarDist) | [10.1007/s11263-021-01639-5](https://doi.org/10.1007/s11263-021-01639-5) |
| SS-B8 | Segmentation | Deep learning automates the quantitative analysis of individual cells in live-cell imaging assays | [10.1371/journal.pcbi.1005177](https://doi.org/10.1371/journal.pcbi.1005177) (DeepCell lineage) |
| SS-B9 | Metadata / interchange | Metadata matters: access to image data in the real world (OME / Bio-Formats framing) | [10.1083/jcb.201004104](https://doi.org/10.1083/jcb.201004104) |
| SS-B10 | Format | OME-Zarr: a cloud-optimized bioimaging file format with international community support | [10.1007/s00418-023-02209-1](https://doi.org/10.1007/s00418-023-02209-1) |
| SS-B11 | Review | Visual interpretability of bioimaging deep learning models | [10.1038/s41592-024-02322-6](https://doi.org/10.1038/s41592-024-02322-6) |

**Use when:** grounding **FORM** §B–E claims, writing Methods, or choosing a default citation for a tool family. **Annotation vs no-label image→counts paths:** §6.

---

## 2. Software registry (quick lookup)

| Tool / project | Role | URL | License / note | Primary cite |
|----------------|------|-----|------------------|--------------|
| Fiji / ImageJ2 | General microscopy workbench | [https://imagej.net/software/fiji](https://imagej.net/software/fiji) | Mixed (plugins vary) | Curate: ImageJ2 paper / Nat Methods ecosystem |
| ImageJ2 core | Next-gen ImageJ | [https://imagej.net/software/imagej2](https://imagej.net/software/imagej2) | BSD-style core | Curate |
| OME Bio-Formats | Parsers for 100+ formats | [https://www.openmicroscopy.org/bio-formats/](https://www.openmicroscopy.org/bio-formats/) | GPL / BSD readers | [10.1083/jcb.201004104](https://doi.org/10.1083/jcb.201004104) (OME framing) |
| CellProfiler | High-throughput cell assays | [https://cellprofiler.org](https://cellprofiler.org) | BSD | [10.1186/gb-2006-7-10-r100](https://doi.org/10.1186/gb-2006-7-10-r100) |
| QuPath | Digital pathology + objects | [https://qupath.github.io](https://qupath.github.io) | GPL v3 | [10.1038/s41598-017-17204-5](https://doi.org/10.1038/s41598-017-17204-5) |
| Ilastik | Pixel / object classification | [https://www.ilastik.org](https://www.ilastik.org) | BSD | [10.1038/nmeth.2082](https://doi.org/10.1038/nmeth.2082) |
| napari | nD viewer + plugin ecosystem | [https://napari.org](https://napari.org) | BSD | [10.1371/journal.pcbi.1008980](https://doi.org/10.1371/journal.pcbi.1008980) |
| Cellpose | Generalist cell segmentation | [https://www.cellpose.org](https://www.cellpose.org) | BSD | [10.1038/s41592-020-01001-x](https://doi.org/10.1038/s41592-020-01001-x) |
| StarDist | Star-convex instance detection | [https://github.com/stardist/stardist](https://github.com/stardist/stardist) | BSD | [10.1007/s11263-021-01639-5](https://doi.org/10.1007/s11263-021-01639-5) |
| DeepCell | Single-cell deep learning toolkit | [https://deepcell.readthedocs.io](https://deepcell.readthedocs.io) | Apache 2.0 | [10.1371/journal.pcbi.1005177](https://doi.org/10.1371/journal.pcbi.1005177) |
| scikit-image | Algorithms in Python | [https://scikit-image.org](https://scikit-image.org) | BSD | [10.7717/peerj.453](https://doi.org/10.7717/peerj.453) |
| OpenCV | General CV primitives | [https://opencv.org](https://opencv.org) | Apache 2.0 | Curate: OpenCV overview paper or docs |
| PyTorch | Deep learning runtime | [https://pytorch.org](https://pytorch.org) | BSD-style | Curate: NeurIPS / foundational cite as needed |
| MorphoLibJ | Mathematical morphology (ImageJ) | [https://imagej.net/plugins/morpholibj](https://imagej.net/plugins/morpholibj) | LGPL | Curate |
| CLIJ / clEsperanto | GPU-accelerated ImageJ ops | [https://clij.github.io](https://clij.github.io) | GPL / MIT components | Curate |
| Labkit | Pixel / instance labeling (ImageJ) | [https://imagej.net/plugins/labkit](https://imagej.net/plugins/labkit) | GPL v2 | Curate |
| CVAT | Team annotation (boxes, masks, tracks) | [https://github.com/cvat-ai/cvat](https://github.com/cvat-ai/cvat) | MIT | Curate |
| Label Studio | Configurable labeling UI | [https://labelstud.io](https://labelstud.io) | Apache 2.0 | Curate |
| VGG VIA | Browser-based image annotation | [https://www.robots.ox.ac.uk/~vgg/software/via/](https://www.robots.ox.ac.uk/~vgg/software/via/) | BSD-2-Clause | Curate |
| μSAM | Segment Anything for Microscopy | [https://github.com/computational-cell-analytics/micro-sam](https://github.com/computational-cell-analytics/micro-sam) | MIT | [10.1038/s41592-024-02580-4](https://doi.org/10.1038/s41592-024-02580-4) |

---

## 3. Public benchmarks and challenge datasets

| Name | What it tests | Entry |
|------|----------------|-------|
| **Broad Bioimage Benchmark Collection (BBBC)** | Classic segmentation / phenotype slots | [https://bbbc.broadinstitute.org](https://bbbc.broadinstitute.org) |
| **Cell Tracking Challenge** | Segmentation + tracking metrics | [http://celltrackingchallenge.net](http://celltrackingchallenge.net) |
| **Human Protein Atlas** (images) | Large-scale IF / tissue context | [https://www.proteinatlas.org](https://www.proteinatlas.org) — Curate: image subset DOI per use case |

**Use when:** closing **FORM** proof steps with numbers comparable to the field.

---

## 3b. StarSight VDM runs (in-vault artifacts)

Internal Vision Dao Monarch (VDM) runs on lab imaging stacks live under `22_Project_StarSight/out/vdm_runs/<run_id>/` as `VDM_REPORT.md`, `vdm-report.json`, and figures (e.g. `figures/luminance_hist.png`). Each run should be logged in [WORLDLINE.md](WORLDLINE.md) with the **EYE-14** tag and dataset description.

- **SS-VDM1** — First full VDM run on current lab imaging dataset (2026-03-31), executed via `scripts/vision_dao_monarch_run.py` inside the `starsight-monarch` Docker EYE. See `out/vdm_runs/20260331T014606Z-*/VDM_REPORT.md` and WORLDLINE Breakthrough 2 for parameters and interpretation.

---

## 4. Cross-domain vision (non-biology) — augmenting the Vision cultivator

**Read with** [FORM.md](FORM.md) §G (“Vision Dao Monarch”). These sources are **not** substitutes for bio tools; they **arm** registration, denoising, QC, and large-tile discipline.

### 4a. Seminal methods (peer-reviewed DOIs)

| ID | Domain | Note | Identifier |
|----|--------|------|------------|
| SS-V1 | Textbook | *Computer Vision: Algorithms and Applications* (2nd ed.) — geometry, restoration, stereo, stitching narrative | [https://szeliski.org/Book/](https://szeliski.org/Book/) (open draft) |
| SS-V10 | Robust fitting | RANSAC — random sample consensus for model fitting with outliers | [10.1145/358669.358692](https://doi.org/10.1145/358669.358692) |
| SS-V11 | Optical flow | Determining optical flow (Horn–Schunck) | [10.1016/0004-3702(81)90024-2](https://doi.org/10.1016/0004-3702(81)90024-2) |
| SS-V12 | Edges | A computational approach to edge detection (Canny) | [10.1109/TPAMI.1986.4767851](https://doi.org/10.1109/TPAMI.1986.4767851) |
| SS-V13 | Registration / flow | Lucas-Kanade 20 Years On: A Unifying Framework | [10.1023/B:VISI.0000011205.11775.fd](https://doi.org/10.1023/B:VISI.0000011205.11775.fd) |
| SS-V14 | Features | Distinctive image features from scale-invariant keypoints (SIFT) | [10.1023/B:VISI.0000029664.99615.94](https://doi.org/10.1023/B:VISI.0000029664.99615.94) |
| SS-V15 | Survey | Image alignment and stitching: a tutorial | [10.1561/0600000009](https://doi.org/10.1561/0600000009) |
| SS-V16 | Denoising | Image denoising by sparse 3-D transform-domain collaborative filtering (BM3D) | [10.1109/TIP.2007.901808](https://doi.org/10.1109/TIP.2007.901808) |
| SS-V17 | Denoising | A non-local algorithm for image denoising | [10.1137/050619618](https://doi.org/10.1137/050619618) |
| SS-V18 | Regularization | Nonlinear total variation based noise removal algorithms (Rudin-Osher-Fatemi) | [10.1016/0167-2789(92)90263-F](https://doi.org/10.1016/0167-2789(92)90263-F) |
| SS-V19 | Quality metric | Image quality assessment: from error visibility to structural similarity (SSIM) | [10.1109/TIP.2003.819861](https://doi.org/10.1109/TIP.2003.819861) |
| SS-V20 | Texture | Textural features for image classification (Haralick) | [10.1109/TSMC.1973.4309314](https://doi.org/10.1109/TSMC.1973.4309314) |
| SS-V21 | Deep backbone | Deep residual learning for image recognition (ResNet) | [10.1109/CVPR.2016.90](https://doi.org/10.1109/CVPR.2016.90) |
| SS-V22 | Deep representation | An image is worth 16x16 words (Vision Transformer) | [openreview.net/forum?id=YicbFdNTTy](https://openreview.net/forum?id=YicbFdNTTy) |

Lucas-Kanade (1981) original: Lucas and Kanade, Proc. Imaging Understanding Workshop -- no single DOI; cite **SS-V13** (Baker-Matthews) or **SS-V15** (Szeliski) in Methods.

### 4b. Software and libraries (general CV / adjacent fields)

| Tool | Role | URL | Primary cite / note |
|------|------|-----|----------------------|
| **OpenCV** | Features, flow, stitching hooks | [https://opencv.org](https://opencv.org) | Curate: official docs + overview paper |
| **Kornia** | Differentiable CV in PyTorch | [https://kornia.github.io](https://kornia.github.io) | Curate |
| **Albumentations** | Augmentations / robustness probes | [https://albumentations.ai](https://albumentations.ai) | Curate |
| **ITK** | nD registration and segmentation | [https://itk.org](https://itk.org) | [10.1016/j.neuroimage.2003.11.055](https://doi.org/10.1016/j.neuroimage.2003.11.055) |
| **VTK** | Visualization pipeline | [https://vtk.org](https://vtk.org) | Curate |
| **GDAL** | Geospatial rasters, tiling, COGs | [https://gdal.org](https://gdal.org) | Curate |
| **Open3D** | 3D geometry / point clouds | [http://www.open3d.org](http://www.open3d.org) | Curate |

### 4c. When non-biology methods help

| Non-biology field | Borrow for microscopy |
|-------------------|------------------------|
| **Robotics / VO** | Drift correction, subpixel alignment |
| **Remote sensing** | Large mosaics, radiometric normalization (mind fluorescence physics) |
| **Industrial inspection** | Anomaly QC maps |
| **Computational photography** | Denoise priors, HDR merge -- disclose processing |

**Use when:** augmenting the **Vision cultivator** per [FORM.md](FORM.md) section G; disclose **domain shift** when priors come from natural images.

---

## 5. Open file formats and interchange

| Format | Role | Pointer |
|--------|------|---------|
| **OME-TIFF** | Metadata + multi-plane in one bundle | [OME documentation](https://docs.openmicroscopy.org/ome-model/latest/ome-tiff/) |
| **OME-Zarr (NGFF)** | Cloud-friendly chunked arrays + metadata | [NGFF spec](https://ngff.openmicroscopy.org/latest/) - [10.1007/s00418-023-02209-1](https://doi.org/10.1007/s00418-023-02209-1) |
| **Plain TIFF / JPEG** | Common exports; may **strip** acquisition metadata -- see **FORM** failure modes | -- |

---

## 6. Annotation tooling and no–manual-label paths (image → counts → analysis)

**Operational map:** [FORM.md](FORM.md) §I. This section inventories **labeling software**, **peer-reviewed methods** that reduce or avoid hand-drawn training masks, and **honest limits** on **cell type** without you annotating biology.

### 6a. Annotation, ROI, and training-data software

| Tool | Role | URL | Primary cite / note |
|------|------|-----|----------------------|
| **QuPath** | Slides, object detection, scoring, optional classifiers | [https://qupath.github.io](https://qupath.github.io) | [10.1038/s41598-017-17204-5](https://doi.org/10.1038/s41598-017-17204-5) |
| **Fiji / ImageJ2** | ROIs, macros, morphology pipelines | [https://imagej.net/software/fiji](https://imagej.net/software/fiji) | Curate |
| **napari** | nD viewing + plugin ecosystem | [https://napari.org](https://napari.org) | [10.1371/journal.pcbi.1008980](https://doi.org/10.1371/journal.pcbi.1008980) |
| **Labkit** | Pixel / instance labels inside ImageJ | [https://imagej.net/plugins/labkit](https://imagej.net/plugins/labkit) | Curate |
| **Ilastik** | Pixel + object classification (user paints examples) | [https://www.ilastik.org](https://www.ilastik.org) | [10.1038/nmeth.2082](https://doi.org/10.1038/nmeth.2082) |
| **CellProfiler** | Module pipelines; often threshold + identify objects | [https://cellprofiler.org](https://cellprofiler.org) | [10.1186/gb-2006-7-10-r100](https://doi.org/10.1186/gb-2006-7-10-r100) |
| **CVAT** | Team labeling (boxes, masks, tracks) | [https://github.com/cvat-ai/cvat](https://github.com/cvat-ai/cvat) | Curate |
| **Label Studio** | Configurable labeling workflows | [https://labelstud.io](https://labelstud.io) | Curate |
| **VGG Image Annotator (VIA)** | Lightweight browser polygons / attributes | [https://www.robots.ox.ac.uk/~vgg/software/via/](https://www.robots.ox.ac.uk/~vgg/software/via/) | Curate |

### 6b. Peer-reviewed: segmentation / counts without *your* hand-drawn masks

These lines still carry **prior knowledge** (pretraining, physics, or architecture)—not magic—but they are the usual citations for “no new pixel labels from this lab.”

| ID | Kind | Title / note | Identifier |
|----|------|--------------|------------|
| SS-NL1 | Unsupervised | UNSEG: unsupervised segmentation of cells and nuclei in complex tissue | [10.1038/s42003-024-06714-4](https://doi.org/10.1038/s42003-024-06714-4) |
| SS-NL2 | Self-supervised | Self-supervised learning for high-throughput / high-content cell segmentation (2025) | [10.1038/s42003-025-08190-w](https://doi.org/10.1038/s42003-025-08190-w) |
| SS-NL3 | Self-supervised | Self-supervised ML for live cell imagery segmentation (motion-based) | [10.1038/s42003-022-04117-x](https://doi.org/10.1038/s42003-022-04117-x) |
| SS-NL4 | Foundation (SAM) | Segment Anything for Microscopy (μSAM) | [10.1038/s41592-024-02580-4](https://doi.org/10.1038/s41592-024-02580-4) |
| SS-NL5 | Foundation | CellSAM: foundation model for cell segmentation | [10.1038/s41592-025-02879-w](https://doi.org/10.1038/s41592-025-02879-w) |
| SS-NL6 | SAM-derived | SAMCell: label-free biological cell segmentation with SAM (PLOS One) | [10.1371/journal.pone.0319532](https://doi.org/10.1371/journal.pone.0319532) |
| SS-NL7 | 3D self-supervised | CellSeg3D: self-supervised 3D cell segmentation | [10.7554/eLife.99848](https://doi.org/10.7554/eLife.99848) |
| SS-NL8 | Embeddings | Cell-DINO: self-supervised embeddings for fluorescence microscopy | [10.1371/journal.pcbi.1013828](https://doi.org/10.1371/journal.pcbi.1013828) |
| SS-NL9 | Preprint | Cell-APP: microscopic cell annotation / segmentation / classification (dataset generation angle) | [10.1101/2025.01.23.634498](https://doi.org/10.1101/2025.01.23.634498) |

**Generalist segmenters (transfer, not “no priors”):** Cellpose [10.1038/s41592-020-01001-x](https://doi.org/10.1038/s41592-020-01001-x), StarDist [10.1007/s11263-021-01639-5](https://doi.org/10.1007/s11263-021-01639-5) — cite **§1** rows SS-B2 / SS-B7 in Methods when used zero-shot on new data.

### 6c. Pipeline lanes: image → instance counts → tables (without drawing masks)

| Lane | Mechanism | Typical outputs | Main caveat |
|------|-----------|-----------------|-------------|
| **A — Classical** | CellProfiler `IdentifyPrimaryObjects`–style threshold + split + size filters; ImageJ morphology | Counts, areas, intensities per object | Fails when foreground/background are not intensity-separable (common on label-free phase/DIC without preprocessing). |
| **B — Pretrained instance** | Cellpose, StarDist, DeepCell on each field | Masks, `n_instances`, morphology stats | **Domain shift**; validate on your modality (see **FORM** §D honest loss). |
| **C — Foundation / SAM family** | μSAM, CellSAM, SAMCell, community wrappers (e.g. PyPI `samcell`) | Masks from prompts or auto-prompting | Zero-shot claims are **dataset-dependent**; still owe QC figures. |
| **D — Unsupervised / SSL** | UNSEG, CellSeg3D, motion-based SSL, etc. | Masks without user labels | Each method assumes **specific data** (e.g. markers, 3D volumes, time-lapse). |
| **E — “Types” without naming** | Embeddings (e.g. Cell-DINO) + clustering | Clusters, UMAPs, silhouette-style QC | Clusters are **not** biological cell types until linked to labels, gating, or orthogonal assays. |

### 6d. Biological cell *type* without exhaustive annotation

Fully **unsupervised** assignment of named phenotypes (e.g. “this cluster is regulatory T cells”) is **not** honest without **some** anchor: intensity gates (QuPath positive detection), **few** reference objects, functional readouts, or expert naming of clusters. Plan Methods accordingly.

### 6e. Vault automation (partial)

| Piece | Role |
|-------|------|
| [`scripts/vision_dao_monarch_run.py`](../scripts/vision_dao_monarch_run.py) | Optional **`--segmentation cellpose`** → per-image `cellpose_n_cells` in `vdm-report.json`; ties to **lane B**. |
| [`requirements-runner-segmentation.txt`](requirements-runner-segmentation.txt) | `torch` + `cellpose` install surface. |
| [FORM.md](FORM.md) §H4 | CLI flags and honesty on **counts vs types**. |

**Use when:** choosing tools for a bounty, writing a no-label Methods paragraph, or extending the Vision Dao Monarch runner toward μSAM / CellProfiler headless.

---

## STARS — US and international anchors

**Public imaging data** alongside §3 (BBBC, Broad) and §5 (OME). US + European archive entry points.

### How to read STARS (context)

**STARS** here are **public image and atlas portals**. They **host or index microscopy**; **cell type, stain, and disease** are in **per-dataset metadata** (and in §§3–5 BBBC / OME notes), not implied by the STAR ID alone.

| ID | What this STAR denotes | Typical use in this BOOK | Not / caveats |
|----|-------------------------|--------------------------|---------------|
| SS-S1 | Cell Image Library (CCDB) | **Curated** cell microscopy collections | Licensing and attribution vary by image set. |
| SS-S2 | BioImage Archive (EMBL-EBI) | **Raw/reference** microscopy datasets | File formats (OME-Zarr, etc.) — follow archive guidance. |
| SS-S3 | Human Protein Atlas (images) | **IF / tissue** context for proteins | Atlas ≠ single-cell RNA; orthogonal modality. |
| SS-S4 | US NLM | **NIH/NLM** library portal | General biomedical resource hub; not imaging-only. |

| ID | Region | Kind | Note | Identifier |
|----|--------|------|------|------------|
| SS-S1 | US | Cell microscopy collection | Cell Image Library (CCDB) | [cellimagelibrary.org](https://www.cellimagelibrary.org/) |
| SS-S2 | UK / EU | Bioimage Archive (EMBL-EBI) | Raw and reference microscopy datasets | [ebi.ac.uk/bioimage-archive](https://www.ebi.ac.uk/bioimage-archive/) |
| SS-S3 | Sweden | Human Protein Atlas (images) | IF / tissue context (see §3 for use) | [proteinatlas.org](https://www.proteinatlas.org/) |
| SS-S4 | US | NIH / NLM biomedical library | National Library of Medicine | [nlm.nih.gov](https://www.nlm.nih.gov/) |

---

BOOK revision: 2026-04-01 — STARS context table.