add a few more nodes

backend/nodes/arc_revolve.py

@@ -0,0 +1,88 @@
+"""Arc Revolve — subtract a cylindrical arc background."""
+
+from __future__ import annotations
+
+import numpy as np
+
+from backend.node_registry import register_node
+from backend.data_types import DataField
+
+
+def _arc_kernel(radius: int) -> np.ndarray:
+    """Build a 1D arc kernel: z = 1 - sqrt(1 - (i/radius)^2)."""
+    half = min(radius, 4096)
+    i = np.arange(-half, half + 1, dtype=np.float64)
+    t = np.clip((i / radius) ** 2, 0.0, 1.0)
+    return 1.0 - np.sqrt(1.0 - t)
+
+
+def _arc_revolve_1d(data: np.ndarray, radius: int) -> np.ndarray:
+    """Compute arc-revolve background for each row independently."""
+    yres, xres = data.shape
+    kernel = _arc_kernel(radius)
+    half = len(kernel) // 2
+    bg = np.empty_like(data)
+
+    for row in range(yres):
+        line = data[row].copy()
+        # Suppress deep outliers before fitting
+        window = min(half, xres // 2)
+        if window > 0:
+            from scipy.ndimage import uniform_filter1d
+            local_mean = uniform_filter1d(line, size=2 * window + 1, mode='nearest')
+            local_sq = uniform_filter1d(line ** 2, size=2 * window + 1, mode='nearest')
+            local_rms = np.sqrt(np.maximum(local_sq - local_mean ** 2, 0.0))
+            threshold = local_mean - 2.5 * np.maximum(local_rms, 1e-30)
+            line = np.maximum(line, threshold)
+
+        # For each pixel, find the lowest position the arc can sit
+        padded = np.pad(line, half, mode='edge')
+        row_bg = np.full(xres, np.inf)
+        for k in range(len(kernel)):
+            shifted = padded[k:k + xres] - kernel[k]
+            row_bg = np.minimum(row_bg, shifted)
+        bg[row] = row_bg
+
+    return bg
+
+
+@register_node(display_name="Arc Revolve")
+class ArcRevolve:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "field": ("DATA_FIELD",),
+                "radius": ("INT", {"default": 20, "min": 1, "max": 1000, "step": 1}),
+                "direction": (["horizontal", "vertical", "both"], {"default": "horizontal"}),
+            }
+        }
+
+    OUTPUTS = (
+        ('DATA_FIELD', 'leveled'),
+        ('DATA_FIELD', 'background'),
+    )
+    FUNCTION = "process"
+
+    DESCRIPTION = (
+        "Subtract a cylindrical arc background. A circular arc of the given "
+        "radius is rolled under each row (or column), and the envelope it "
+        "traces out is subtracted as the background."
+    )
+
+    KEYWORDS = ("arc", "revolve", "cylindrical", "background", "level")
+
+    def process(self, field: DataField, radius: int = 20,
+                direction: str = "horizontal") -> tuple:
+        data = np.asarray(field.data, dtype=np.float64)
+
+        if direction == "horizontal":
+            bg = _arc_revolve_1d(data, radius)
+        elif direction == "vertical":
+            bg = _arc_revolve_1d(data.T, radius).T
+        else:
+            bg_h = _arc_revolve_1d(data, radius)
+            bg_v = _arc_revolve_1d(data.T, radius).T
+            bg = np.minimum(bg_h, bg_v)
+
+        return (field.replace(data=data - bg), field.replace(data=bg))

backend/nodes/level_rotate.py

@@ -0,0 +1,75 @@
+"""Level Rotate — level by physically rotating the data plane."""
+
+from __future__ import annotations
+
+import numpy as np
+from scipy.ndimage import map_coordinates
+
+from backend.node_registry import register_node
+from backend.data_types import DataField
+
+
+@register_node(display_name="Level Rotate")
+class LevelRotate:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "field": ("DATA_FIELD",),
+            }
+        }
+
+    OUTPUTS = (
+        ('DATA_FIELD', 'leveled'),
+    )
+    FUNCTION = "process"
+
+    DESCRIPTION = (
+        "Level by physically rotating the data plane. Fits a best-fit plane, "
+        "converts its slopes to tilt angles, then rotates the surface by "
+        "those angles using interpolation rather than algebraic subtraction."
+    )
+
+    KEYWORDS = ("rotate", "tilt", "level", "plane")
+
+    def process(self, field: DataField) -> tuple:
+        data = np.asarray(field.data, dtype=np.float64)
+        yres, xres = data.shape
+
+        # Fit plane: z = a + bx*x + by*y (x,y in pixel coords)
+        yy, xx = np.mgrid[:yres, :xres].astype(np.float64)
+        A = np.column_stack([np.ones(yres * xres), xx.ravel(), yy.ravel()])
+        coeffs, _, _, _ = np.linalg.lstsq(A, data.ravel(), rcond=None)
+        _, bx, by = coeffs
+
+        # Convert pixel slopes to tilt angles
+        alpha_x = np.arctan(bx)
+        alpha_y = np.arctan(by)
+
+        # Build rotation: for each output pixel, find where it came from
+        cx = (xres - 1) / 2.0
+        cy = (yres - 1) / 2.0
+
+        cos_x = np.cos(alpha_x)
+        cos_y = np.cos(alpha_y)
+
+        # Source coordinates after removing tilt
+        src_x = xx.copy()
+        src_y = yy.copy()
+        src_z = data.copy()
+
+        # Rotate about x-axis (corrects y-tilt)
+        dy = yy - cy
+        src_y_rot = cy + dy * cos_y
+        src_z = src_z - dy * np.sin(alpha_y)
+
+        # Rotate about y-axis (corrects x-tilt)
+        dx = xx - cx
+        src_x_rot = cx + dx * cos_x
+        src_z = src_z - dx * np.sin(alpha_x)
+
+        # Resample with the adjusted z values
+        result = map_coordinates(src_z, [src_y_rot, src_x_rot], order=1,
+                                 mode='nearest')
+
+        return (field.replace(data=result),)

backend/nodes/sphere_revolve.py

@@ -0,0 +1,74 @@
+"""Sphere Revolve — subtract a spherical cap background."""
+
+from __future__ import annotations
+
+import numpy as np
+from scipy.ndimage import uniform_filter
+
+from backend.node_registry import register_node
+from backend.data_types import DataField
+
+
+def _sphere_kernel(radius: int) -> np.ndarray:
+    """Build a 2D spherical cap kernel."""
+    half = min(radius, 512)
+    i = np.arange(-half, half + 1, dtype=np.float64)
+    ii, jj = np.meshgrid(i, i)
+    r2 = (ii ** 2 + jj ** 2) / (radius ** 2)
+    r2 = np.clip(r2, 0.0, 1.0)
+    return 1.0 - np.sqrt(1.0 - r2)
+
+
+@register_node(display_name="Sphere Revolve")
+class SphereRevolve:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "field": ("DATA_FIELD",),
+                "radius": ("INT", {"default": 20, "min": 1, "max": 500, "step": 1}),
+            }
+        }
+
+    OUTPUTS = (
+        ('DATA_FIELD', 'leveled'),
+        ('DATA_FIELD', 'background'),
+    )
+    FUNCTION = "process"
+
+    DESCRIPTION = (
+        "Subtract a spherical cap background. A sphere of the given radius "
+        "is rolled under the surface, and the envelope it traces is "
+        "subtracted as the background."
+    )
+
+    KEYWORDS = ("sphere", "revolve", "spherical", "background", "level")
+
+    def process(self, field: DataField, radius: int = 20) -> tuple:
+        data = np.asarray(field.data, dtype=np.float64)
+        yres, xres = data.shape
+
+        kernel = _sphere_kernel(radius)
+        half = kernel.shape[0] // 2
+
+        # Suppress deep outliers
+        window = max(1, half // 2)
+        local_mean = uniform_filter(data, size=2 * window + 1, mode='nearest')
+        local_sq = uniform_filter(data ** 2, size=2 * window + 1, mode='nearest')
+        local_rms = np.sqrt(np.maximum(local_sq - local_mean ** 2, 0.0))
+        threshold = local_mean - 2.5 * np.maximum(local_rms, 1e-30)
+        clipped = np.maximum(data, threshold)
+
+        padded = np.pad(clipped, half, mode='edge')
+        bg = np.full_like(data, np.inf)
+
+        ks = kernel.shape[0]
+        for di in range(ks):
+            for dj in range(ks):
+                k_val = kernel[di, dj]
+                if k_val >= 1.0:
+                    continue
+                shifted = padded[di:di + yres, dj:dj + xres] - k_val
+                bg = np.minimum(bg, shifted)
+
+        return (field.replace(data=data - bg), field.replace(data=bg))

backend/nodes/unrotate.py

@@ -0,0 +1,88 @@
+"""Unrotate — auto-detect and correct in-plane scan rotation."""
+
+from __future__ import annotations
+
+import numpy as np
+from scipy.ndimage import rotate as ndimage_rotate
+
+from backend.node_registry import register_node
+from backend.data_types import DataField
+
+
+def _slope_angle_histogram(data: np.ndarray, n_bins: int = 3600) -> np.ndarray:
+    """Compute histogram of local slope angles over [0, 2*pi)."""
+    dy = np.diff(data, axis=0)[:, :-1]
+    dx = np.diff(data, axis=1)[:-1, :]
+    angles = np.arctan2(dy, dx) % (2 * np.pi)
+    hist, _ = np.histogram(angles.ravel(), bins=n_bins, range=(0, 2 * np.pi))
+    return hist.astype(np.float64)
+
+
+def _find_dominant_angle(hist: np.ndarray, symmetry: int) -> float:
+    """Find the rotation correction angle for a given symmetry order.
+
+    Folds the histogram into one symmetry sector, finds the peak, and
+    returns the offset to the nearest axis.
+    """
+    n_bins = len(hist)
+    sector = n_bins // symmetry
+    folded = np.zeros(sector, dtype=np.float64)
+    for k in range(symmetry):
+        start = k * sector
+        end = start + sector
+        if end <= n_bins:
+            folded += hist[start:end]
+
+    peak_bin = int(np.argmax(folded))
+    bin_angle = (2 * np.pi / symmetry) / sector
+
+    # The angle of the peak
+    peak_angle = peak_bin * bin_angle
+
+    # The nearest axis is at multiples of pi/symmetry
+    axis_spacing = np.pi / symmetry
+    nearest_axis = round(peak_angle / axis_spacing) * axis_spacing
+    correction = nearest_axis - peak_angle
+
+    return float(correction)
+
+
+@register_node(display_name="Unrotate")
+class Unrotate:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "field": ("DATA_FIELD",),
+                "symmetry": (["2-fold", "3-fold", "4-fold", "6-fold"], {"default": "4-fold"}),
+            }
+        }
+
+    OUTPUTS = (
+        ('DATA_FIELD', 'leveled'),
+    )
+    FUNCTION = "process"
+
+    DESCRIPTION = (
+        "Auto-detect and correct in-plane scan rotation. Computes a slope "
+        "angle histogram, finds the dominant feature direction for the given "
+        "symmetry, and rotates the image to align features with the axes."
+    )
+
+    KEYWORDS = ("rotation", "alignment", "angle", "symmetry", "crystal")
+
+    def process(self, field: DataField, symmetry: str = "4-fold") -> tuple:
+        data = np.asarray(field.data, dtype=np.float64)
+
+        sym_order = int(symmetry[0])
+        hist = _slope_angle_histogram(data)
+        angle_rad = _find_dominant_angle(hist, sym_order)
+        angle_deg = float(np.degrees(angle_rad))
+
+        if abs(angle_deg) < 0.01:
+            return (field,)
+
+        rotated = ndimage_rotate(data, angle_deg, reshape=False, order=1,
+                                 mode='nearest')
+
+        return (field.replace(data=rotated),)

backend/nodes/zero_value.py

@@ -0,0 +1,56 @@
+"""Zero Value — shift data so the mean or maximum equals zero."""
+
+from __future__ import annotations
+
+import numpy as np
+
+from backend.node_registry import register_node
+from backend.data_types import DataField
+
+
+@register_node(display_name="Zero Mean")
+class ZeroMean:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "field": ("DATA_FIELD",),
+            }
+        }
+
+    OUTPUTS = (
+        ('DATA_FIELD', 'leveled'),
+    )
+    FUNCTION = "process"
+
+    DESCRIPTION = "Shift all values so the mean is exactly zero."
+
+    KEYWORDS = ("offset", "center", "level", "mean")
+
+    def process(self, field: DataField) -> tuple:
+        data = np.asarray(field.data, dtype=np.float64)
+        return (field.replace(data=data - data.mean()),)
+
+
+@register_node(display_name="Zero Maximum")
+class ZeroMaximum:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "field": ("DATA_FIELD",),
+            }
+        }
+
+    OUTPUTS = (
+        ('DATA_FIELD', 'leveled'),
+    )
+    FUNCTION = "process"
+
+    DESCRIPTION = "Shift all values so the maximum is exactly zero."
+
+    KEYWORDS = ("offset", "level", "maximum")
+
+    def process(self, field: DataField) -> tuple:
+        data = np.asarray(field.data, dtype=np.float64)
+        return (field.replace(data=data - data.max()),)