Added anti-aliasing to Threshold node (#2209)

backend/src/packages/chaiNNer_standard/__init__.py

@@ -70,7 +70,7 @@
         Dependency(
             display_name="ChaiNNer Extensions",
             pypi_name="chainner_ext",
-            version="0.3.0",
+            version="0.3.1",
             size_estimate=2.0 * MB,
         ),
     ],

backend/src/packages/chaiNNer_standard/image_adjustment/adjustments/threshold.py

@@ -5,9 +5,10 @@
 
 import cv2
 import numpy as np
+from chainner_ext import binary_threshold
 
 from nodes.groups import if_enum_group
-from nodes.properties.inputs import EnumInput, ImageInput, SliderInput
+from nodes.properties.inputs import BoolInput, EnumInput, ImageInput, SliderInput
 from nodes.properties.outputs import ImageOutput
 
 from .. import adjustments_group
@@ -59,6 +60,10 @@ class ThresholdType(Enum):
                 controls_step=1,
             ).with_id(2),
         ),
+        BoolInput("Anti-aliasing", default=False).with_docs(
+            "Enables sub-pixel precision. Bilinear interpolation is used to fill in values in between pixels.",
+            "Conceptually, the option is equivalent to first upscaling the image by a factor of X (with linear interpolation), thresholding it, and then downscaling it by a factor of X (where X is 20 or more).",
+        ),
     ],
     outputs=[ImageOutput(image_type="Input0")],
     see_also=[
@@ -71,10 +76,26 @@ def threshold_node(
     threshold: float,
     thresh_type: ThresholdType,
     max_value: float,
+    anti_aliasing: bool,
 ) -> np.ndarray:
     threshold /= 100
     max_value /= 100
 
-    _, result = cv2.threshold(img, threshold, max_value, thresh_type.value)
+    if not anti_aliasing:
+        _, result = cv2.threshold(img, threshold, max_value, thresh_type.value)
+        return result
+
+    binary = binary_threshold(img, threshold, True)
+    if thresh_type == ThresholdType.BINARY_INV:
+        binary = 1 - binary
 
-    return result
+    if thresh_type == ThresholdType.BINARY or thresh_type == ThresholdType.BINARY_INV:
+        if max_value < 1:
+            binary *= max_value
+        return binary
+    elif thresh_type == ThresholdType.TRUNC:
+        return binary * threshold + img * (1 - binary)
+    elif thresh_type == ThresholdType.TO_ZERO:
+        return binary * img
+    elif thresh_type == ThresholdType.TO_ZERO_INV:
+        return (1 - binary) * img