diff --git a/engine/pipeline/preprocessing.py b/engine/pipeline/preprocessing.py
new file mode 100644
index 0000000..fddc037
--- /dev/null
+++ b/engine/pipeline/preprocessing.py
@@ -0,0 +1,130 @@
+"""OpenCV preprocessing pipeline for raster-to-vector conversion."""
+
+import cv2
+import numpy as np
+
+
+def decode_image(raw_bytes: bytes) -> np.ndarray:
+    """Decode raw image bytes into a BGR numpy array."""
+    buf = np.frombuffer(raw_bytes, dtype=np.uint8)
+    img = cv2.imdecode(buf, cv2.IMREAD_COLOR)
+    if img is None:
+        raise ValueError("Failed to decode image from provided bytes")
+    return img
+
+
+def to_grayscale(img: np.ndarray) -> np.ndarray:
+    """Convert BGR image to single-channel grayscale."""
+    if len(img.shape) == 2:
+        return img
+    return cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+
+
+def denoise(
+    img: np.ndarray,
+    d: int = 9,
+    sigma_color: float = 75.0,
+    sigma_space: float = 75.0,
+) -> np.ndarray:
+    """Apply bilateral filter for edge-preserving denoising."""
+    return cv2.bilateralFilter(img, d, sigma_color, sigma_space)
+
+
+def enhance_contrast(
+    img: np.ndarray,
+    clip_limit: float = 2.0,
+    tile_grid_size: tuple[int, int] = (8, 8),
+) -> np.ndarray:
+    """Apply CLAHE contrast enhancement."""
+    clahe = cv2.createCLAHE(clipLimit=clip_limit, tileGridSize=tile_grid_size)
+    return clahe.apply(img)
+
+
+def threshold(
+    img: np.ndarray,
+    manual_thresh: int | None = None,
+) -> np.ndarray:
+    """Apply thresholding — Otsu auto by default, manual override if provided."""
+    if manual_thresh is not None:
+        _, result = cv2.threshold(img, manual_thresh, 255, cv2.THRESH_BINARY)
+    else:
+        _, result = cv2.threshold(img, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
+    return result
+
+
+def edge_detect(
+    img: np.ndarray,
+    low: int = 50,
+    high: int = 150,
+) -> np.ndarray:
+    """Apply Canny edge detection."""
+    return cv2.Canny(img, low, high)
+
+
+def morphological_ops(
+    img: np.ndarray,
+    kernel_size: int = 3,
+    dilate_iterations: int = 1,
+    erode_iterations: int = 1,
+) -> np.ndarray:
+    """Apply dilation then erosion (closing-style) to clean up binary image."""
+    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (kernel_size, kernel_size))
+    result = cv2.dilate(img, kernel, iterations=dilate_iterations)
+    result = cv2.erode(result, kernel, iterations=erode_iterations)
+    return result
+
+
+def preprocess(
+    raw_bytes: bytes,
+    params: dict | None = None,
+) -> np.ndarray:
+    """Run the full preprocessing pipeline on raw image bytes.
+
+    Stages: decode → grayscale → denoise → contrast → threshold → morphological ops.
+    Edge detection is optional (enabled via params["edge_detect"] = True).
+
+    All stage parameters are tunable via the params dict. Keys:
+        denoise_d, denoise_sigma_color, denoise_sigma_space,
+        clahe_clip_limit, clahe_tile_grid_size,
+        threshold_manual,
+        edge_detect (bool), edge_low, edge_high,
+        morph_kernel_size, morph_dilate_iterations, morph_erode_iterations
+    """
+    p = params or {}
+
+    img = decode_image(raw_bytes)
+    img = to_grayscale(img)
+
+    img = denoise(
+        img,
+        d=p.get("denoise_d", 9),
+        sigma_color=p.get("denoise_sigma_color", 75.0),
+        sigma_space=p.get("denoise_sigma_space", 75.0),
+    )
+
+    img = enhance_contrast(
+        img,
+        clip_limit=p.get("clahe_clip_limit", 2.0),
+        tile_grid_size=p.get("clahe_tile_grid_size", (8, 8)),
+    )
+
+    img = threshold(
+        img,
+        manual_thresh=p.get("threshold_manual"),
+    )
+
+    if p.get("edge_detect", False):
+        img = edge_detect(
+            img,
+            low=p.get("edge_low", 50),
+            high=p.get("edge_high", 150),
+        )
+
+    img = morphological_ops(
+        img,
+        kernel_size=p.get("morph_kernel_size", 3),
+        dilate_iterations=p.get("morph_dilate_iterations", 1),
+        erode_iterations=p.get("morph_erode_iterations", 1),
+    )
+
+    return img
diff --git a/engine/tests/test_preprocessing.py b/engine/tests/test_preprocessing.py
new file mode 100644
index 0000000..366d33b
--- /dev/null
+++ b/engine/tests/test_preprocessing.py
@@ -0,0 +1,232 @@
+"""Tests for the OpenCV preprocessing pipeline."""
+
+import cv2
+import numpy as np
+import pytest
+
+from pipeline.preprocessing import (
+    decode_image,
+    denoise,
+    edge_detect,
+    enhance_contrast,
+    morphological_ops,
+    preprocess,
+    threshold,
+    to_grayscale,
+)
+
+
+def _make_test_image(width: int = 100, height: int = 80, color: bool = True) -> np.ndarray:
+    """Create a synthetic test image with some structure."""
+    if color:
+        img = np.zeros((height, width, 3), dtype=np.uint8)
+        # White rectangle on black background
+        cv2.rectangle(img, (20, 15), (80, 65), (255, 255, 255), -1)
+        # Gray circle for tonal variation
+        cv2.circle(img, (50, 40), 15, (128, 128, 128), -1)
+    else:
+        img = np.zeros((height, width), dtype=np.uint8)
+        cv2.rectangle(img, (20, 15), (80, 65), 255, -1)
+        cv2.circle(img, (50, 40), 15, 128, -1)
+    return img
+
+
+def _encode_png(img: np.ndarray) -> bytes:
+    """Encode numpy array to PNG bytes."""
+    ok, buf = cv2.imencode(".png", img)
+    assert ok
+    return buf.tobytes()
+
+
+# --- decode_image ---
+
+class TestDecodeImage:
+    def test_decodes_valid_png(self):
+        img = _make_test_image()
+        raw = _encode_png(img)
+        result = decode_image(raw)
+        assert result.shape == img.shape
+        assert result.dtype == np.uint8
+
+    def test_rejects_invalid_bytes(self):
+        with pytest.raises(ValueError, match="Failed to decode"):
+            decode_image(b"not an image")
+
+    def test_decodes_jpeg(self):
+        img = _make_test_image()
+        ok, buf = cv2.imencode(".jpg", img)
+        assert ok
+        result = decode_image(buf.tobytes())
+        assert result.shape == img.shape
+
+
+# --- to_grayscale ---
+
+class TestToGrayscale:
+    def test_converts_color_to_gray(self):
+        img = _make_test_image(color=True)
+        result = to_grayscale(img)
+        assert len(result.shape) == 2
+        assert result.shape == (80, 100)
+
+    def test_passthrough_already_gray(self):
+        img = _make_test_image(color=False)
+        result = to_grayscale(img)
+        assert len(result.shape) == 2
+        np.testing.assert_array_equal(result, img)
+
+
+# --- denoise ---
+
+class TestDenoise:
+    def test_returns_same_shape(self):
+        img = _make_test_image(color=False)
+        result = denoise(img)
+        assert result.shape == img.shape
+        assert result.dtype == np.uint8
+
+    def test_custom_params(self):
+        img = _make_test_image(color=False)
+        result = denoise(img, d=5, sigma_color=50.0, sigma_space=50.0)
+        assert result.shape == img.shape
+
+    def test_reduces_noise(self):
+        img = _make_test_image(color=False).astype(np.float64)
+        rng = np.random.default_rng(42)
+        noisy = np.clip(img + rng.normal(0, 25, img.shape), 0, 255).astype(np.uint8)
+        result = denoise(noisy, d=9, sigma_color=75, sigma_space=75)
+        # Denoised image should be closer to original than noisy version
+        orig = _make_test_image(color=False)
+        noise_diff = np.mean(np.abs(noisy.astype(float) - orig.astype(float)))
+        clean_diff = np.mean(np.abs(result.astype(float) - orig.astype(float)))
+        assert clean_diff < noise_diff
+
+
+# --- enhance_contrast ---
+
+class TestEnhanceContrast:
+    def test_returns_same_shape(self):
+        img = _make_test_image(color=False)
+        result = enhance_contrast(img)
+        assert result.shape == img.shape
+
+    def test_custom_clip_limit(self):
+        img = _make_test_image(color=False)
+        result = enhance_contrast(img, clip_limit=4.0, tile_grid_size=(4, 4))
+        assert result.shape == img.shape
+
+    def test_increases_dynamic_range(self):
+        # Low-contrast gray image
+        img = np.full((80, 100), 120, dtype=np.uint8)
+        img[20:60, 20:80] = 130
+        result = enhance_contrast(img, clip_limit=2.0)
+        # CLAHE should increase the spread
+        assert result.std() >= img.std()
+
+
+# --- threshold ---
+
+class TestThreshold:
+    def test_otsu_produces_binary(self):
+        img = _make_test_image(color=False)
+        result = threshold(img)
+        unique = set(np.unique(result))
+        assert unique <= {0, 255}
+
+    def test_manual_threshold(self):
+        img = _make_test_image(color=False)
+        result = threshold(img, manual_thresh=100)
+        unique = set(np.unique(result))
+        assert unique <= {0, 255}
+
+    def test_manual_threshold_value(self):
+        img = np.array([[50, 150], [100, 200]], dtype=np.uint8)
+        result = threshold(img, manual_thresh=120)
+        expected = np.array([[0, 255], [0, 255]], dtype=np.uint8)
+        np.testing.assert_array_equal(result, expected)
+
+
+# --- edge_detect ---
+
+class TestEdgeDetect:
+    def test_returns_same_shape(self):
+        img = _make_test_image(color=False)
+        result = edge_detect(img)
+        assert result.shape == img.shape
+
+    def test_detects_edges(self):
+        img = np.zeros((100, 100), dtype=np.uint8)
+        img[25:75, 25:75] = 255
+        result = edge_detect(img, low=50, high=150)
+        # Should have non-zero pixels along the rectangle edges
+        assert np.count_nonzero(result) > 0
+        # Interior and exterior should be zero
+        assert result[50, 50] == 0
+        assert result[0, 0] == 0
+
+    def test_custom_thresholds(self):
+        img = _make_test_image(color=False)
+        result = edge_detect(img, low=100, high=200)
+        assert result.dtype == np.uint8
+
+
+# --- morphological_ops ---
+
+class TestMorphologicalOps:
+    def test_returns_same_shape(self):
+        img = threshold(_make_test_image(color=False))
+        result = morphological_ops(img)
+        assert result.shape == img.shape
+
+    def test_fills_small_gaps(self):
+        # Create binary image with a small hole
+        img = np.zeros((100, 100), dtype=np.uint8)
+        img[20:80, 20:80] = 255
+        img[49:51, 49:51] = 0  # small gap
+        result = morphological_ops(img, kernel_size=3, dilate_iterations=1, erode_iterations=1)
+        # Gap should be filled after dilation
+        assert result[49, 49] == 255
+
+    def test_custom_kernel_and_iterations(self):
+        img = threshold(_make_test_image(color=False))
+        result = morphological_ops(img, kernel_size=5, dilate_iterations=2, erode_iterations=2)
+        assert result.shape == img.shape
+
+
+# --- full pipeline ---
+
+class TestPreprocess:
+    def test_end_to_end_default_params(self):
+        img = _make_test_image()
+        raw = _encode_png(img)
+        result = preprocess(raw)
+        assert len(result.shape) == 2
+        assert result.dtype == np.uint8
+        unique = set(np.unique(result))
+        # After threshold + morph, should be mostly binary
+        assert unique <= {0, 255}
+
+    def test_with_edge_detection(self):
+        img = _make_test_image()
+        raw = _encode_png(img)
+        result = preprocess(raw, params={"edge_detect": True})
+        assert len(result.shape) == 2
+
+    def test_custom_params(self):
+        img = _make_test_image()
+        raw = _encode_png(img)
+        result = preprocess(raw, params={
+            "denoise_d": 5,
+            "denoise_sigma_color": 50.0,
+            "clahe_clip_limit": 4.0,
+            "threshold_manual": 128,
+            "morph_kernel_size": 5,
+            "morph_dilate_iterations": 2,
+            "morph_erode_iterations": 2,
+        })
+        assert result.dtype == np.uint8
+        assert len(result.shape) == 2
+
+    def test_rejects_bad_input(self):
+        with pytest.raises(ValueError):
+            preprocess(b"garbage data")