From b569b035721fded90ba754c914891e42502e0eae Mon Sep 17 00:00:00 2001 From: llocarnini Date: Sat, 5 Feb 2022 18:00:10 +0100 Subject: [PATCH] current status of trying to connect layout and table parsing --- table_parsing/table_parsig.py | 94 +++++++++++++++++++++++------------ 1 file changed, 63 insertions(+), 31 deletions(-) diff --git a/table_parsing/table_parsig.py b/table_parsing/table_parsig.py index 16db709..00a3d36 100644 --- a/table_parsing/table_parsig.py +++ b/table_parsing/table_parsig.py @@ -9,19 +9,22 @@ from matplotlib import pyplot as plt def parse(image: np.array): gray_scale = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) - blurred = cv2.GaussianBlur(gray_scale, (5, 5), 1) - thresh = cv2.threshold(blurred, 253, 255, cv2.THRESH_BINARY)[1] + #plt.imshow(gray_scale) + blurred = cv2.GaussianBlur(gray_scale, (7, 7), 2) #5 5 1 + thresh = cv2.threshold(blurred, 251, 255, cv2.THRESH_BINARY)[1] + #plt.imshow(thresh) img_bin = ~thresh - line_min_width = 10 + line_min_width = 7 kernel_h = np.ones((10, line_min_width), np.uint8) kernel_v = np.ones((line_min_width, 10), np.uint8) img_bin_h = cv2.morphologyEx(img_bin, cv2.MORPH_OPEN, kernel_h) img_bin_v = cv2.morphologyEx(img_bin, cv2.MORPH_OPEN, kernel_v) - + #plt.imshow(img_bin_h) + #plt.imshow(img_bin_v) img_bin_final = img_bin_h | img_bin_v - + plt.imshow(img_bin_final) contours = cv2.findContours(img_bin_final, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) contours = imutils.grab_contours(contours) for c in contours: @@ -29,25 +32,48 @@ def parse(image: np.array): approx = cv2.approxPolyDP(c, 0.04 * peri, True) yield cv2.boundingRect(approx) -def parse_tables(image: np.array, rectangle): - gray_scale = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) - th1, img_bin = cv2.threshold(gray_scale, 200, 255, cv2.THRESH_BINARY) - img_bin = ~img_bin +def parse_tables(image: np.array, rects: list): + parsed_tables = [] + for rect in rects: + (x,y,w,h) = rect + region_of_interest = image[x:x+w, y:y+h] + gray = cv2.cvtColor(region_of_interest, cv2.COLOR_BGR2GRAY) + thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY)[1] + img_bin = ~thresh - line_min_width = 5 - kernel_h = np.ones((1, line_min_width), np.uint8) - kernel_v = np.ones((line_min_width, 1), np.uint8) + line_min_width = 5 + kernel_h = np.ones((1, line_min_width), np.uint8) + kernel_v = np.ones((line_min_width, 1), np.uint8) - img_bin_h = cv2.morphologyEx(img_bin, cv2.MORPH_OPEN, kernel_h) - img_bin_v = cv2.morphologyEx(img_bin, cv2.MORPH_OPEN, kernel_v) + img_bin_h = cv2.morphologyEx(img_bin, cv2.MORPH_OPEN, kernel_h) + img_bin_v = cv2.morphologyEx(img_bin, cv2.MORPH_OPEN, kernel_v) # find_and_close_internal_gaps(img_bin_v) - img_bin_final = img_bin_h | img_bin_v - plt.imshow(img_bin_final) + img_bin_final = img_bin_h | img_bin_v + #plt.imshow(img_bin_final) # find_and_close_internal_gaps(img_bin_final) # find_and_close_edges(img_bin_final) - _, labels, stats, _ = cv2.connectedComponentsWithStats(~img_bin_final, connectivity=8, ltype=cv2.CV_32S) - return labels, stats + _, labels, stats, _ = cv2.connectedComponentsWithStats(~img_bin_final, connectivity=8, ltype=cv2.CV_32S) + parsed_tables.append([(x,y,w,h), stats]) + return parsed_tables + #yield (x,y,w,h), stats, region_of_interest + # return stats + +def annotate_table(image, parsed_tables): + for table in parsed_tables: + original_coordinates, stats = table + stats = filter_unconnected_cells(stats) + for stat in stats: + x, y, w, h, area = stat + cv2.rectangle(image, (x, y), (x + w, y + h), (255, 0, 255), 2) + for i, (s, v) in enumerate(zip(["x", "y", "w", "h"], [x, y, w, h])): + anno = f"{s} = {v}" + xann = int(x + 5) + yann = int(y + h - (20 * (i + 1))) + cv2.putText(image, anno, (xann, yann), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255, 0, 255), 2) + + return image + def filter_unconnected_cells(stats): @@ -87,18 +113,7 @@ def find_and_close_edges(img_bin_final): return img_bin_final -def annotate_image(image, stats): - stats = filter_unconnected_cells(stats) - for stat in stats: - x, y, w, h, area = stat - cv2.rectangle(image, (x, y), (x + w, y + h), (255, 0, 255), 2) - for i, (s, v) in enumerate(zip(["x", "y", "w", "h"], [x, y, w, h])): - anno = f"{s} = {v}" - xann = int(x + 5) - yann = int(y + h - (20 * (i + 1))) - cv2.putText(image, anno, (xann, yann), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255, 0, 255), 2) - return image def parse_tables_in_pdf(pages): return zip(map(parse, pages), count()) @@ -118,19 +133,36 @@ def parse_tables_in_pdf(pages): def annotate_boxes(image, rects): + print(type(rects)) for rect in rects: (x, y, w, h) = rect cv2.rectangle(image, (x, y), (x + w, y + h), (0, 255, 0), 2) return image +def filter_tables_or_images(rects): + filtered = [] + for rect in rects: + (x,y,w,h) = rect + print(w*h) + if w * h > 10**6: + filtered.append(rect) + print(filtered) + return filtered + + + def annotate_tables_in_pdf(pdf_path, page_index=1): page = pdf2image.convert_from_path(pdf_path, first_page=page_index + 1, last_page=page_index + 1)[0] page = np.array(page) - asd = parse(page) - page = annotate_boxes(page, asd) + layout_boxes = parse(page) + page = annotate_boxes(page, layout_boxes) + parsed_tables = parse_tables(page, filter_tables_or_images(layout_boxes)) + page = annotate_table(page, parsed_tables) + + fig, ax = plt.subplots(1, 1) fig.set_size_inches(20, 20)