140 lines
5.6 KiB
Python
140 lines
5.6 KiB
Python
"""Perspective correction — fix perspective distortion using a projective transform."""
|
|
|
|
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, datafield_to_uint8, encode_preview
|
|
from backend.execution_context import emit_overlay
|
|
|
|
|
|
@register_node(display_name="Perspective Correction")
|
|
class PerspectiveCorrection:
|
|
_CUSTOM_PREVIEW = True
|
|
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"field": ("DATA_FIELD",),
|
|
"top_left_x": ("FLOAT", {"default": 0.1, "min": -1.0, "max": 1.0, "step": 0.01, "hidden": True}),
|
|
"top_left_y": ("FLOAT", {"default": 0.1, "min": -1.0, "max": 1.0, "step": 0.01, "hidden": True}),
|
|
"top_right_x": ("FLOAT", {"default": -0.1, "min": -1.0, "max": 1.0, "step": 0.01, "hidden": True}),
|
|
"top_right_y": ("FLOAT", {"default": 0.1, "min": -1.0, "max": 1.0, "step": 0.01, "hidden": True}),
|
|
"bottom_left_x": ("FLOAT", {"default": 0.1, "min": -1.0, "max": 1.0, "step": 0.01, "hidden": True}),
|
|
"bottom_left_y": ("FLOAT", {"default": -0.1, "min": -1.0, "max": 1.0, "step": 0.01, "hidden": True}),
|
|
"bottom_right_x": ("FLOAT", {"default": -0.1, "min": -1.0, "max": 1.0, "step": 0.01, "hidden": True}),
|
|
"bottom_right_y": ("FLOAT", {"default": -0.1, "min": -1.0, "max": 1.0, "step": 0.01, "hidden": True}),
|
|
},
|
|
"optional": {
|
|
"top_left": ("COORD",),
|
|
"top_right": ("COORD",),
|
|
"bottom_left": ("COORD",),
|
|
"bottom_right": ("COORD",),
|
|
},
|
|
}
|
|
|
|
OUTPUTS = (
|
|
('DATA_FIELD', 'corrected'),
|
|
)
|
|
FUNCTION = "process"
|
|
|
|
DESCRIPTION = (
|
|
"Fix perspective distortion by dragging corner handles. Each corner "
|
|
"offset defines a distorted quadrilateral that is warped back to "
|
|
"a rectangle."
|
|
)
|
|
|
|
KEYWORDS = ("keystone", "homography", "projective", "warp", "quadrilateral", "distortion")
|
|
|
|
def process(self, field: DataField,
|
|
top_left_x: float, top_left_y: float,
|
|
top_right_x: float, top_right_y: float,
|
|
bottom_left_x: float, bottom_left_y: float,
|
|
bottom_right_x: float, bottom_right_y: float,
|
|
top_left: tuple[float, float] | None = None,
|
|
top_right: tuple[float, float] | None = None,
|
|
bottom_left: tuple[float, float] | None = None,
|
|
bottom_right: tuple[float, float] | None = None) -> tuple:
|
|
if top_left is not None:
|
|
top_left_x, top_left_y = float(top_left[0]), float(top_left[1])
|
|
if top_right is not None:
|
|
top_right_x, top_right_y = float(top_right[0]), float(top_right[1])
|
|
if bottom_left is not None:
|
|
bottom_left_x, bottom_left_y = float(bottom_left[0]), float(bottom_left[1])
|
|
if bottom_right is not None:
|
|
bottom_right_x, bottom_right_y = float(bottom_right[0]), float(bottom_right[1])
|
|
|
|
data = np.asarray(field.data, dtype=np.float64)
|
|
yres, xres = data.shape
|
|
|
|
src = np.array([
|
|
[top_left_y * yres, top_left_x * xres],
|
|
[top_right_y * yres, top_right_x * xres + (xres - 1)],
|
|
[(1 + bottom_left_y) * yres - 1, bottom_left_x * xres],
|
|
[(1 + bottom_right_y) * yres - 1, bottom_right_x * xres + (xres - 1)],
|
|
], dtype=np.float64)
|
|
|
|
dst = np.array([
|
|
[0, 0],
|
|
[0, xres - 1],
|
|
[yres - 1, 0],
|
|
[yres - 1, xres - 1],
|
|
], dtype=np.float64)
|
|
|
|
H = _solve_perspective(src, dst)
|
|
|
|
yy, xx = np.mgrid[:yres, :xres]
|
|
coords = np.stack([xx.ravel(), yy.ravel(), np.ones(yres * xres)])
|
|
src_coords = H @ coords
|
|
src_coords /= src_coords[2:3, :]
|
|
sx = src_coords[0].reshape(yres, xres)
|
|
sy = src_coords[1].reshape(yres, xres)
|
|
|
|
result = map_coordinates(data, [sy, sx], order=1, mode='nearest')
|
|
corrected = field.replace(data=result)
|
|
|
|
source_rgb = datafield_to_uint8(field, field.colormap)
|
|
corrected_rgb = datafield_to_uint8(corrected, corrected.colormap)
|
|
|
|
corners = [
|
|
{"x": float(top_left_x), "y": float(top_left_y)},
|
|
{"x": float(top_right_x), "y": float(top_right_y)},
|
|
{"x": float(bottom_left_x), "y": float(bottom_left_y)},
|
|
{"x": float(bottom_right_x), "y": float(bottom_right_y)},
|
|
]
|
|
|
|
emit_overlay({
|
|
"kind": "perspective",
|
|
"section_title": "Perspective",
|
|
"image": encode_preview(source_rgb),
|
|
"corrected_image": encode_preview(corrected_rgb),
|
|
"corners": corners,
|
|
})
|
|
|
|
return (corrected,)
|
|
|
|
|
|
def _solve_perspective(src: np.ndarray, dst: np.ndarray) -> np.ndarray:
|
|
"""Solve for 3x3 perspective matrix mapping dst -> src (for inverse warp).
|
|
|
|
Coordinates are (col, row) — the matrix is applied to [col, row, 1] vectors.
|
|
"""
|
|
n = len(src)
|
|
A = np.zeros((2 * n, 8))
|
|
b = np.zeros(2 * n)
|
|
for i in range(n):
|
|
dr, dc = dst[i] # dest row, col
|
|
sr, sc = src[i] # src row, col
|
|
A[2 * i] = [dc, dr, 1, 0, 0, 0, -sc * dc, -sc * dr]
|
|
A[2 * i + 1] = [0, 0, 0, dc, dr, 1, -sr * dc, -sr * dr]
|
|
b[2 * i] = sc
|
|
b[2 * i + 1] = sr
|
|
h, _, _, _ = np.linalg.lstsq(A, b, rcond=None)
|
|
H = np.array([[h[0], h[1], h[2]],
|
|
[h[3], h[4], h[5]],
|
|
[h[6], h[7], 1.0]])
|
|
return H
|