Add script to ocr individual cells

pdf_table_extraction_and_ocr.org

@@ -500,9 +500,7 @@ thick noise or a wrongly detected table.
 If everything looks reasonable but the OCR is doing something like turning a
 period into a comma, then you might need to do some custom Tesseract training.
 
-#+BEGIN_SRC python :noweb-ref ocr-image :eval no :noweb yes
-<<crop-to-text>>
-
+#+BEGIN_SRC python :noweb-ref ocr-image :eval no
 def ocr_image(image, config):
     cropped = crop_to_text(image)
     return pytesseract.image_to_string(
@@ -515,6 +513,7 @@ def ocr_image(image, config):
 import pytesseract
 import cv2
 image = cv2.imread("resources/examples/example-table-cell-1-1.png", cv2.IMREAD_GRAYSCALE)
+<<crop-to-text>>
 <<ocr-image>>
 ocr_image(image, "--psm 7")
 #+END_SRC
@@ -733,6 +732,35 @@ if __name__ == "__main__":
     main(sys.argv[1])
 #+END_SRC
 
+*** table_ocr/ocr_image.py
+
+This does a little bit of cleanup before sending it through tesseract.
+
+#+BEGIN_SRC shell :results output
+. ~/.virtualenvs/lotto_odds/bin/activate
+python -m table_ocr.ocr_cell resources/examples/cells/000-000.png
+#+END_SRC
+
+#+RESULTS:
+: PRIZE
+
+#+BEGIN_SRC python :tangle table_ocr/ocr_cell.py :mkdirp yes :results none
+import sys
+
+import cv2
+import pytesseract
+
+<<crop-to-text>>
+<<ocr-image>>
+
+def main(f):
+    image = cv2.imread(f, cv2.IMREAD_GRAYSCALE)
+    print(ocr_image(image, "--psm 7"))
+
+if __name__ == "__main__":
+    main(sys.argv[1])
+#+END_SRC
+
 * Utils
 
 The following code lets us specify a size for images when they are exported to

table_ocr/extract_tables.py

@@ -13,7 +13,6 @@ def main(files):
         directory, filename = os.path.split(f)
 
         image = cv2.imread(f, cv2.IMREAD_GRAYSCALE)
-        print("Reading {}".format(f))
         tables = find_tables(image)
         files = []
         for i, table in enumerate(tables):

table_ocr/ocr_cell.py

@@ -0,0 +1,30 @@
+import sys
+
+import cv2
+import pytesseract
+
+def crop_to_text(image):
+    kernel = cv2.getStructuringElement(cv2.MORPH_CROSS, (4, 4))
+    opened = cv2.morphologyEx(~image, cv2.MORPH_OPEN, kernel)
+
+    contours, hierarchy = cv2.findContours(opened, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
+    bounding_rects = [cv2.boundingRect(c) for c in contours]
+    # The largest contour is certainly the text that we're looking for.
+    largest_rect = max(bounding_rects, key=lambda r: r[2] * r[3])
+    x, y, w, h = largest_rect
+    cropped = image[y:y+h, x:x+w]
+    bordered = cv2.copyMakeBorder(cropped, 5, 5, 5, 5, cv2.BORDER_CONSTANT, None, 255)
+    return bordered
+def ocr_image(image, config):
+    cropped = crop_to_text(image)
+    return pytesseract.image_to_string(
+        ~cropped,
+        config=config
+    )
+
+def main(f):
+    image = cv2.imread(f, cv2.IMREAD_GRAYSCALE)
+    print(ocr_image(image, "--psm 7"))
+
+if __name__ == "__main__":
+    main(sys.argv[1])