Merge pull request #740 from sbrunner/auto-mask

Auto mask

Makefile

@@ -2,6 +2,7 @@ export DOCKER_BUILDKIT=1
 
 .PHONY: jsonschema
 jsonschema:
+	ci/schema-copy
 	jsonschema2md scan_to_paperless/config_schema.json config.md
 	jsonschema2md scan_to_paperless/process_schema.json process.md
 	c2cciutils-checks --fix --check=prettier

ci/schema-copy

@@ -0,0 +1,18 @@
+#!/usr/bin/env python3
+
+import json
+
+
+def _main() -> None:
+    with open("scan_to_paperless/config_schema.json", encoding="utf-8") as config_schema_file:
+        config_schema = json.load(config_schema_file)
+    with open("scan_to_paperless/process_schema.json", encoding="utf-8") as process_schema_file:
+        process_schema = json.load(process_schema_file)
+
+    process_schema["definitions"]["args"]["properties"] = config_schema["definitions"]["args"]["properties"]
+    with open("scan_to_paperless/process_schema.json", "w", encoding="utf-8") as process_schema_file:
+        json.dump(process_schema, process_schema_file, indent=2)
+
+
+if __name__ == "__main__":
+    _main()

config.md

@@ -58,3 +58,12 @@
   - **`run_ps2pdf`** _(boolean)_: Run the ps2pdf optimizer (=> JPEG). Default: `False`.
   - **`jpeg`** _(boolean)_: Convert images to JPEG. Default: `False`.
   - **`jpeg_quality`** _(integer)_: The JPEG quality. Default: `90`.
+  - **`auto_mask`** _(object)_: The auto mask configuration.
+    - **`lower_hsv_color`** _(array)_: The lower color in HSV representation. Default: `[0, 0, 108]`.
+      - **Items** _(integer)_
+    - **`upper_hsv_color`** _(array)_: The upper color in HSV representation. Default: `[255, 10, 148]`.
+      - **Items** _(integer)_
+    - **`de_noise_size`** _(integer)_: The size of the artifact that will be de noise. Default: `20`.
+    - **`de_noise_level`** _(integer)_: The threshold level used in de noise on the blurry image. Default: `220`.
+    - **`buffer_size`** _(integer)_: The size of the buffer add on the mask. Default: `100`.
+    - **`buffer_level`** _(integer)_: The threshold level used in buffer on the blurry image. Default: `20`.

process.md

@@ -87,3 +87,12 @@
   - **`run_ps2pdf`** _(boolean)_: Run the ps2pdf optimizer (=> JPEG). Default: `False`.
   - **`jpeg`** _(boolean)_: Convert images to JPEG. Default: `False`.
   - **`jpeg_quality`** _(integer)_: The JPEG quality. Default: `90`.
+  - **`auto_mask`** _(object)_: The auto mask configuration.
+    - **`lower_hsv_color`** _(array)_: The lower color in HSV representation. Default: `[0, 0, 108]`.
+      - **Items** _(integer)_
+    - **`upper_hsv_color`** _(array)_: The upper color in HSV representation. Default: `[255, 10, 148]`.
+      - **Items** _(integer)_
+    - **`de_noise_size`** _(integer)_: The size of the artifact that will be de noise. Default: `20`.
+    - **`de_noise_level`** _(integer)_: The threshold level used in de noise on the blurry image. Default: `220`.
+    - **`buffer_size`** _(integer)_: The size of the buffer add on the mask. Default: `100`.
+    - **`buffer_level`** _(integer)_: The threshold level used in buffer on the blurry image. Default: `20`.

scan_to_paperless/config.py

@@ -152,6 +152,9 @@
         #
         # default: 90
         "jpeg_quality": int,
+        # WARNING: The required are not correctly taken in account,
+        # See: https://github.com/camptocamp/jsonschema-gentypes/issues/6
+        "auto_mask": "_ArgumentsAutoMask",
     },
     total=False,
 )
@@ -192,6 +195,45 @@
 )
 
 
+# The auto mask configuration
+_ArgumentsAutoMask = TypedDict(
+    "_ArgumentsAutoMask",
+    {
+        # The lower color in HSV representation
+        #
+        # default:
+        #   - 0
+        #   - 0
+        #   - 108
+        "lower_hsv_color": List[int],
+        # The upper color in HSV representation
+        #
+        # default:
+        #   - 255
+        #   - 10
+        #   - 148
+        "upper_hsv_color": List[int],
+        # The size of the artifact that will be de noise
+        #
+        # default: 20
+        "de_noise_size": int,
+        # The threshold level used in de noise on the blurry image
+        #
+        # default: 220
+        "de_noise_level": int,
+        # The size of the buffer add on the mask
+        #
+        # default: 100
+        "buffer_size": int,
+        # The threshold level used in buffer on the blurry image
+        #
+        # default: 20
+        "buffer_level": int,
+    },
+    total=False,
+)
+
+
 _ConfigurationModesAdditionalproperties = TypedDict(
     "_ConfigurationModesAdditionalproperties",
     {

scan_to_paperless/config_schema.json

@@ -198,6 +198,48 @@
           "type": "integer",
           "default": 90,
           "description": "The JPEG quality"
+        },
+        "auto_mask": {
+          "type": "object",
+          "description": "The auto mask configuration",
+          "properties": {
+            "lower_hsv_color": {
+              "type": "array",
+              "description": "The lower color in HSV representation",
+              "default": [0, 0, 108],
+              "items": {
+                "type": "integer"
+              }
+            },
+            "upper_hsv_color": {
+              "type": "array",
+              "description": "The upper color in HSV representation",
+              "default": [255, 10, 148],
+              "items": {
+                "type": "integer"
+              }
+            },
+            "de_noise_size": {
+              "type": "integer",
+              "description": "The size of the artifact that will be de noise",
+              "default": 20
+            },
+            "de_noise_level": {
+              "type": "integer",
+              "description": "The threshold level used in de noise on the blurry image",
+              "default": 220
+            },
+            "buffer_size": {
+              "type": "integer",
+              "description": "The size of the buffer add on the mask",
+              "default": 100
+            },
+            "buffer_level": {
+              "type": "integer",
+              "description": "The threshold level used in buffer on the blurry image",
+              "default": 20
+            }
+          }
         }
       }
     }

scan_to_paperless/process.py

@@ -102,13 +102,57 @@ def __init__(  # pylint: disable=too-many-arguments
 
     def init_mask(self) -> None:
         """Init the mask."""
-        if self.image is None:
-            raise Exception("The image is None")
-        if self.mask is None:
-            raise Exception("The mask is None")
-        self.mask_ready = cv2.resize(
-            cv2.cvtColor(self.mask, cv2.COLOR_BGR2GRAY), (self.image.shape[1], self.image.shape[0])
-        )
+        if "auto_mask" in self.config["args"]:
+            auto_mask_config = self.config["args"]["auto_mask"]
+            hsv = cv2.cvtColor(self.image, cv2.COLOR_BGR2HSV)
+
+            lower_val = np.array(auto_mask_config.get("lower_hsv_color", [0, 0, 108]))
+            upper_val = np.array(auto_mask_config.get("upper_hsv_color", [255, 10, 148]))
+            mask = cv2.inRange(hsv, lower_val, upper_val)
+            de_noise_size = auto_mask_config.get("de_noise_size", 20)
+            mask = cv2.copyMakeBorder(
+                mask,
+                de_noise_size,
+                de_noise_size,
+                de_noise_size,
+                de_noise_size,
+                cv2.BORDER_CONSTANT,
+                value=255,
+            )
+
+            blur = cv2.blur(
+                mask,
+                (de_noise_size, de_noise_size),
+            )
+            _, thresh1 = cv2.threshold(
+                blur, auto_mask_config.get("de_noise_level", 220), 255, cv2.THRESH_BINARY
+            )
+
+            blur = cv2.blur(
+                thresh1,
+                (
+                    auto_mask_config.get("buffer_size", 100),
+                    auto_mask_config.get("buffer_size", 100),
+                ),
+            )
+            _, mask = cv2.threshold(blur, auto_mask_config.get("buffer_level", 20), 255, cv2.THRESH_BINARY)
+            self.mask = 255 - mask[de_noise_size:-de_noise_size, de_noise_size:-de_noise_size]
+            if os.environ.get("PROGRESS", "FALSE") == "TRUE" and self.root_folder:
+                cv2.imwrite(os.path.join(self.root_folder, "mask.png"), self.mask)
+        elif self.root_folder:
+            mask_file = os.path.join(self.root_folder, "mask.png")
+            if not os.path.exists(mask_file):
+                base_folder = os.path.dirname(self.root_folder)
+                assert base_folder
+                mask_file = os.path.join(base_folder, "mask.png")
+                if not os.path.exists(mask_file):
+                    return
+
+            self.mask = cv2.imread(mask_file)
+
+        if self.image is not None and self.mask is not None:
+            maskbw = self.mask if len(self.mask.shape) == 2 else cv2.cvtColor(self.mask, cv2.COLOR_BGR2GRAY)
+            self.mask_ready = cv2.resize(maskbw, (self.image.shape[1], self.image.shape[0]))
 
     def get_process_count(self) -> int:
         """Get the step number."""
@@ -230,9 +274,11 @@ def image_diff(image1: NpNdarrayInt, image2: NpNdarrayInt) -> Tuple[float, NpNda
     height = max(image1.shape[0], image2.shape[0])
     image1 = cv2.resize(image1, (width, height))
     image2 = cv2.resize(image2, (width, height))
-    score, diff = structural_similarity(
-        cv2.cvtColor(image1, cv2.COLOR_BGR2GRAY), cv2.cvtColor(image2, cv2.COLOR_BGR2GRAY), full=True
-    )
+
+    image1 = image1 if len(image1.shape) == 2 else cv2.cvtColor(image1, cv2.COLOR_BGR2GRAY)
+    image2 = image2 if len(image2.shape) == 2 else cv2.cvtColor(image2, cv2.COLOR_BGR2GRAY)
+
+    score, diff = structural_similarity(image1, image2, full=True)
     diff = (255 - diff * 255).astype("uint8")
     return score, diff
 
@@ -812,10 +858,7 @@ def transform(
         image_status = image_config.setdefault("status", {})
         assert context.image is not None
         image_status["size"] = list(context.image.shape[:2][::-1])
-        mask_file = os.path.join(os.path.dirname(root_folder), "mask.png")
-        if os.path.exists(mask_file):
-            context.mask = cv2.imread(mask_file)
-            context.init_mask()
+        context.init_mask()
         level(context)
         deskew(context)
         docrop(context)

scan_to_paperless/process_schema.json

@@ -6,6 +6,7 @@
   "additionalProperties": false,
   "definitions": {
     "args": {
+      "Editor note": "The properties of this object should be modified in the config_schema.json file",
       "type": "object",
       "title": "Arguments",
       "properties": {
@@ -176,7 +177,9 @@
         "pngquant_options": {
           "type": "array",
           "description": "The pngquant options",
-          "items": { "type": "string" }
+          "items": {
+            "type": "string"
+          }
         },
         "run_exiftool": {
           "type": "boolean",
@@ -197,14 +200,58 @@
           "type": "integer",
           "default": 90,
           "description": "The JPEG quality"
+        },
+        "auto_mask": {
+          "type": "object",
+          "description": "The auto mask configuration",
+          "properties": {
+            "lower_hsv_color": {
+              "type": "array",
+              "description": "The lower color in HSV representation",
+              "default": [0, 0, 108],
+              "items": {
+                "type": "integer"
+              }
+            },
+            "upper_hsv_color": {
+              "type": "array",
+              "description": "The upper color in HSV representation",
+              "default": [255, 10, 148],
+              "items": {
+                "type": "integer"
+              }
+            },
+            "de_noise_size": {
+              "type": "integer",
+              "description": "The size of the artifact that will be de noise",
+              "default": 20
+            },
+            "de_noise_level": {
+              "type": "integer",
+              "description": "The threshold level used in de noise on the blurry image",
+              "default": 220
+            },
+            "buffer_size": {
+              "type": "integer",
+              "description": "The size of the buffer add on the mask",
+              "default": 100
+            },
+            "buffer_level": {
+              "type": "integer",
+              "description": "The threshold level used in buffer on the blurry image",
+              "default": 20
+            }
+          }
         }
       }
     }
   },
   "properties": {
     "images": {
       "type": "array",
-      "items": { "type": "string" },
+      "items": {
+        "type": "string"
+      },
       "description": "The images"
     },
     "args": {
@@ -235,7 +282,9 @@
           "sources": {
             "type": "array",
             "description": "The images obtain after the current step",
-            "items": { "type": "string" }
+            "items": {
+              "type": "string"
+            }
           },
           "process_count": {
             "type": "integer",
@@ -248,7 +297,6 @@
       "type": "array",
       "items": {
         "title": "Assisted split",
-
         "type": "object",
         "additionalProperties": false,
         "description": "Assisted split configuration",