diff --git a/.gitignore b/.gitignore
index a3ea72a..b96d5ee 100644
--- a/.gitignore
+++ b/.gitignore
@@ -42,6 +42,7 @@ coverage.xml
 *.svg
 *.png
 *.pdf
+!docs/_static/images/*.png
 *_checkpoint.ipynb
 
 
diff --git a/docs/_static/images/overall.png b/docs/_static/images/overall.png
new file mode 100644
index 0000000..d690035
Binary files /dev/null and b/docs/_static/images/overall.png differ
diff --git a/docs/index.rst b/docs/index.rst
index db8d2a0..51faabf 100644
--- a/docs/index.rst
+++ b/docs/index.rst
@@ -8,7 +8,7 @@ multi-stain whole-slide images (WSIs).  It combines prompt-based tissue masking,
 alignment, and nuclei-level fine registration to deliver precise cell-level correspondence across
 stains.
 
-.. figure:: _static/images/pipeline_overview.svg
+.. figure:: _static/images/overall.png
    :alt: CORE pipeline overview
    :align: center
    :width: 100%
diff --git a/docs/quickstart.rst b/docs/quickstart.rst
index 5e9d18e..93a02a7 100644
--- a/docs/quickstart.rst
+++ b/docs/quickstart.rst
@@ -4,14 +4,12 @@ Quick Start
 The fastest way to get started is to run the end-to-end notebooks inside
 the ``notebooks/`` folder.  The steps below show the minimal Python API.
 
-.. figure:: _static/images/coarse_fine_registration.svg
-   :alt: Coarse-to-fine registration strategy
+.. figure:: _static/images/overall.png
+   :alt: CORE pipeline overview
    :align: center
    :width: 100%
 
-   **Figure 1.** CORE's two-stage registration strategy.  The coarse stage performs rigid and
-   elastic alignment at low resolution; the fine stage refines cell-level correspondence using
-   nuclei point-sets and shape-aware coherent point drift (CPD).
+   **Figure 1.** CORE pipeline overview used as an illustrative reference in this quick start.
 
 1. Edit ``config.py``
 ---------------------
@@ -48,14 +46,6 @@ Set at least the two WSI paths and the desired resolutions:
 4. Extract tissue masks
 -----------------------
 
-.. figure:: _static/images/tissue_mask_example.svg
-   :alt: Prompt-based tissue masking
-   :align: center
-   :width: 95%
-
-   **Figure 2.** Prompt-based tissue segmentation.  VisionAgent uses a text prompt to isolate
-   the tissue region, producing a binary mask that is then used to guide registration.
-
 .. code-block:: python
 
    from core.preprocessing.preprocessing import extract_tissue_masks
@@ -118,15 +108,6 @@ nuclei CSV files with ``global_x`` and ``global_y`` columns.
 8. Evaluate
 -----------
 
-.. figure:: _static/images/evaluation_tre.svg
-   :alt: Target Registration Error before and after registration
-   :align: center
-   :width: 95%
-
-   **Figure 3.** Target Registration Error (TRE) before and after CORE registration.  Blue dots
-   are fixed landmark positions; crosses show the corresponding moving landmarks.  A good
-   registration brings the two sets of points into close agreement.
-
 .. code-block:: python
 
    from core.evaluation.evaluation import evaluate_registration_tre