Pass wrong_angle_split_line.png

bansan85 · Nov 16, 2020 · ce033e5 · ce033e5
1 parent 5b14280
commit ce033e5
Show file tree

Hide file tree

Showing 7 changed files with 258 additions and 162 deletions.
diff --git a/cv2ext.py b/cv2ext.py
@@ -136,17 +136,9 @@ def remove_border_in_contours(
     height, width = get_hw(image)
     for cnt in contours:
         for contour in cnt:
-            if contour[0, 0] - border_size < -1:
-                raise ValueError("Contour should not be inside the border.")
-            if contour[0, 0] - border_size > width:
-                raise ValueError("Contour should not be inside the border.")
             contour[0, 0] = compute.clamp(
                 contour[0, 0] - border_size, 0, width - 1
             )
-            if contour[0, 1] - border_size < -1:
-                raise ValueError("Contour should not be inside the border.")
-            if contour[0, 1] - border_size > height:
-                raise ValueError("Contour should not be inside the border.")
             contour[0, 1] = compute.clamp(
                 contour[0, 1] - border_size, 0, height - 1
             )
@@ -290,6 +282,7 @@ def __insert_border_in_mask(
     epsilon: Tuple[int, float],
     page_angle: float,
 ) -> None:
+    __pourcentage_white_allowed__ = 0.01
     epsilon_border, epsilon_angle = epsilon
     height, width = get_hw(threshold2)
     cnt2 = cnt[cnt[:, 0, 0] > epsilon_border]
@@ -377,9 +370,11 @@ def __insert_border_in_mask(
             )
         mask = cv2.drawContours(mask, [pts], 0, 255, -1)
         histogram = cv2.calcHist([threshold2], [0], mask, [2], [0, 256])
-        if 0.1 * histogram[0] > sum(histogram[1:]) or 0.1 * histogram[
-            -1
-        ] > sum(histogram[:-1]):
+        if __pourcentage_white_allowed__ * histogram[0] > sum(
+            histogram[1:]
+        ) or __pourcentage_white_allowed__ * histogram[-1] > sum(
+            histogram[:-1]
+        ):
             mask_border_only = cv2.drawContours(
                 mask_border_only, [pts], 0, (0), -1
             )
@@ -408,9 +403,11 @@ def __insert_border_in_mask(
             )
         mask = cv2.drawContours(mask, [pts], 0, 255, -1)
         histogram = cv2.calcHist([threshold2], [0], mask, [2], [0, 256])
-        if 0.1 * histogram[0] > sum(histogram[1:]) or 0.1 * histogram[
-            -1
-        ] > sum(histogram[:-1]):
+        if __pourcentage_white_allowed__ * histogram[0] > sum(
+            histogram[1:]
+        ) or __pourcentage_white_allowed__ * histogram[-1] > sum(
+            histogram[:-1]
+        ):
             mask_border_only = cv2.drawContours(
                 mask_border_only, [pts], 0, (0), -1
             )
@@ -419,14 +416,13 @@ def __insert_border_in_mask(
 def remove_black_border_in_image(
     gray_bordered: Any, page_angle: float, enable_debug: Optional[str]
 ) -> Any:
-    gray_bordered2 = cv2.bitwise_not(gray_bordered)
-    write_image_if(gray_bordered2, enable_debug, "_2a.png")
-    _, threshold = cv2.threshold(gray_bordered, 15, 255, cv2.THRESH_BINARY)
-    write_image_if(threshold, enable_debug, "_2b.png")
-    threshold2 = cv2.bitwise_not(threshold)
-    write_image_if(threshold2, enable_debug, "_2b2.png")
+    thresholdi = threshold_from_gaussian_histogram_black(gray_bordered)
+    _, threshold = cv2.threshold(
+        gray_bordered, thresholdi, 255, cv2.THRESH_BINARY_INV
+    )
+    write_image_if(threshold, enable_debug, "_2b2.png")
     contours, _ = cv2.findContours(
-        threshold2, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE
+        threshold, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE
     )
     if enable_debug is not None:
         image_contours = cv2.drawContours(
@@ -453,53 +449,88 @@ def remove_black_border_in_image(
         ):
             __insert_border_in_mask(
                 cnt,
-                threshold2,
+                threshold,
                 mask_border_only,
                 (__epsilon__, __angle_tolerance__),
                 page_angle,
             )
 
-    # Borders are in black.
+    # Borders are in black in mask.
     write_image_if(mask_border_only, enable_debug, "_2c.png")
-    res = cv2.bitwise_and(
-        gray_bordered2, gray_bordered2, mask=mask_border_only
+    return mask_border_only
+
+
+def apply_mask(image: Any, mask: Any) -> Any:
+    gray_bordered2 = cv2.bitwise_not(image)
+    gray_bordered3 = cv2.bitwise_and(
+        gray_bordered2, gray_bordered2, mask=mask
     )
-    write_image_if(res, enable_debug, "_2d.png")
-    gray_bordered2 = cv2.bitwise_not(res)
-    write_image_if(gray_bordered2, enable_debug, "_2e.png")
-    return gray_bordered2
+    gray_bordered4 = cv2.bitwise_not(gray_bordered3)
+    # Borders are in white in original image.
+    return gray_bordered4
 
 
 def erode_and_dilate(
-    image: Any, size: Tuple[int, int], iterations: int
+    image: Any, size: Tuple[int, int], iterations: int, reverse: bool = False
 ) -> Any:
-    eroded = cv2.erode(
-        image,
-        cv2.getStructuringElement(cv2.MORPH_ELLIPSE, size),
-        iterations=iterations,
-    )
-    dilate = cv2.dilate(
-        eroded,
-        cv2.getStructuringElement(cv2.MORPH_ELLIPSE, size),
-        iterations=iterations,
-    )
-    return dilate
+    start = int(reverse)
+    img = image
+    for i in range(2):
+        if (i + start) % 2 == 0:
+            img = cv2.erode(
+                img,
+                cv2.getStructuringElement(cv2.MORPH_ELLIPSE, size),
+                iterations=iterations,
+            )
+        else:
+            img = cv2.dilate(
+                img,
+                cv2.getStructuringElement(cv2.MORPH_ELLIPSE, size),
+                iterations=iterations,
+            )
+    return img
 
 
-def threshold_from_gaussian_histogram(
-    image: Any, pourcentage: float = 0.8
-) -> Any:
+def threshold_from_gaussian_histogram_white(
+    image: Any, pourcentage: float = 0.2, blur_kernel_size: int = 31
+) -> int:
     histogram = cv2.calcHist([image], [0], None, [256], [0, 256])
+    histogram_blur = cv2.GaussianBlur(
+        histogram,
+        (1, blur_kernel_size),
+        blur_kernel_size,
+        borderType=cv2.BORDER_REPLICATE,
+    )
     i = 255
-    for j in range(255, 0, -1):
-        if histogram[j][0] == 0:
+    extreme_min = histogram_blur[255][0]
+    for j in range(254, 0, -1):
+        if histogram_blur[j][0] < extreme_min:
+            extreme_min = histogram_blur[j][0]
+        else:
             i = j
             break
+    limit = extreme_min * (1 + pourcentage)
     for j in range(i, 0, -1):
-        if histogram[j][0] != 0:
+        if histogram_blur[j][0] > limit:
             i = j
             break
-    return int(pourcentage * i)
+    return i
+
+
+def threshold_from_gaussian_histogram_black(
+    image: Any, blur_kernel_size: int = 31
+) -> int:
+    histogram = cv2.calcHist([image], [0], None, [256], [0, 256])
+    histogram_blur = cv2.GaussianBlur(
+        histogram,
+        (1, blur_kernel_size),
+        blur_kernel_size,
+        borderType=cv2.BORDER_REPLICATE,
+    )
+    for i in range(1, 256):
+        if histogram_blur[i][0] < histogram_blur[i + 1][0]:
+            return i
+    return 255
 
 
 def gaussian_blur_wrap(histogram: Any, kernel_size: int) -> Any:
@@ -637,3 +668,19 @@ def bounding_rectangle(
             [[rectangle[0] + rectangle[2], rectangle[1]]],
         ]
     )
+
+
+def is_line_not_cross_images(
+    line: Tuple[int, int, int, int], images_mask: Any
+) -> bool:
+    line_x1, line_y1, line_x2, line_y2 = line
+    image_line = np.zeros(images_mask.shape, np.uint8)
+    cv2.line(
+        image_line,
+        (line_x1, line_y1),
+        (line_x2, line_y2),
+        (255, 255, 255),
+        1,
+    )
+    image_line = cv2.bitwise_and(images_mask, image_line)
+    return cv2.countNonZero(image_line) == 0
diff --git a/page/crop.py b/page/crop.py
@@ -58,11 +58,10 @@ class Impl(types.SimpleNamespace):
             1, np.pi / (180 * 20), 30, 100, 30
         )
         find_images: FindImageParameters = FindImageParameters(
-            0.005,
+            5,
             (10, 10),
             (10, 10),
             (10, 10),
-            0.005,
             0.01,
         )
 
@@ -130,11 +129,13 @@ class CropAroundDataInPageParameters:
     class Impl(types.SimpleNamespace):
         found_data_try1: FoundDataTry1Parameters = FoundDataTry1Parameters()
         found_data_try2: FoundDataTry2Parameters = FoundDataTry2Parameters()
-        dilate_size: Tuple[int, int] = (2, 2)
+        dilate_size: Tuple[int, int] = (5, 5)
         threshold2: int = 200
-        contour_area_min: float = 0.002 * 0.002
-        contour_area_max: float = 0.5 * 0.5
+        contour_area_min: float = 0.01 * 0.01
+        contour_area_max: float = 1.0
         border: int = 10
+        skip_rectangle_closed_to_line: float = 5.0
+        closed_to_edge: float = 0.02
 
     def __init__(self) -> None:
         self.__param = CropAroundDataInPageParameters.Impl()
@@ -191,6 +192,22 @@ def border(self) -> int:
     def border(self, val: int) -> None:
         self.__param.border = val
 
+    @property
+    def skip_rectangle_closed_to_line(self) -> float:
+        return self.__param.skip_rectangle_closed_to_line
+
+    @skip_rectangle_closed_to_line.setter
+    def skip_rectangle_closed_to_line(self, val: float) -> None:
+        self.__param.skip_rectangle_closed_to_line = val
+
+    @property
+    def closed_to_edge(self) -> float:
+        return self.__param.closed_to_edge
+
+    @closed_to_edge.setter
+    def closed_to_edge(self, val: float) -> None:
+        self.__param.closed_to_edge = val
+
     def init_default_values(
         self,
         key: str,
@@ -711,9 +728,8 @@ def crop_around_data(
 
     gray = cv2ext.convertion_en_niveau_de_gris(page_gauche_0)
 
-    dilated = cv2.dilate(
-        gray,
-        cv2.getStructuringElement(cv2.MORPH_ELLIPSE, parameters.dilate_size),
+    dilated = cv2ext.erode_and_dilate(
+        gray, parameters.dilate_size, parameters.dilate_size[0]
     )
     cv2ext.write_image_if(dilated, enable_debug, "_" + str(n_page) + "_7.png")
 
@@ -726,19 +742,53 @@ def crop_around_data(
     cv2ext.write_image_if(
         threshold, enable_debug, "_" + str(n_page) + "_8.png"
     )
-    contours, _ = cv2.findContours(
-        threshold, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE
+    threshold2 = cv2.copyMakeBorder(
+        threshold,
+        1,
+        1,
+        1,
+        1,
+        cv2.BORDER_CONSTANT,
+        value=[255],
     )
+    contours, hierarchy = cv2.findContours(
+        threshold2, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE
+    )
+    cv2ext.remove_border_in_contours(contours, 1, threshold)
     if enable_debug is not None:
         image2222 = cv2ext.convertion_en_couleur(page_gauche_0)
     ncontour_good_size = False
+    first_cnt_all = int(cv2.contourArea(contours[0])) == (imgh - 1) * (
+        imgw - 1
+    )
+
+    def is_border(contour: Any) -> bool:
+        rectangle = cv2.boundingRect(contour)
+        ratio = rectangle[3] / rectangle[2]
+        if ratio > parameters.skip_rectangle_closed_to_line and (
+            rectangle[0] < parameters.closed_to_edge * imgw
+            or rectangle[0] > (1 - parameters.closed_to_edge) * imgw
+        ):
+            return True
+        if ratio < 1 / parameters.skip_rectangle_closed_to_line and (
+            rectangle[1] < parameters.closed_to_edge * imgh
+            or rectangle[1] > (1 - parameters.closed_to_edge) * imgh
+        ):
+            return True
+        return False
+
     contours_listered = filter(
         lambda x: parameters.contour_area_min * imgh * imgw
-        < cv2.contourArea(x)
-        < parameters.contour_area_max * imgh * imgw,
-        contours,
+        < cv2.contourArea(x[0])
+        < parameters.contour_area_max * imgh * imgw
+        and (
+            (x[1][3] == -1 and not first_cnt_all)
+            or (x[1][3] == 0 and first_cnt_all)
+        )
+        and not is_border(x[0]),
+        zip(contours, hierarchy[0]),
     )
-    for cnt in contours_listered:
+    for cnt, _ in contours_listered:
         (point_x, point_y, width, height) = cv2.boundingRect(cnt)
         if enable_debug is not None:
             cv2.drawContours(image2222, [cnt], -1, (0, 0, 255), 3)