def test_image_text_mask_with_east(testdata_dir):
img = CompareImage(testdata_dir / 'Beach_date.png',
placeholder_file=testdata_dir / 'pattern_mask.json',
ocr_engine='east')
assert len(img.placeholders) >= 1
img_with_mask = img.get_image_with_placeholders()
assert img != img_with_mask
def get_text_from_document(self, image):
"""Gets Text Content from documents/images ``image``.
Text content is returned as a list of strings. None if no text is identified.
Examples:
| ${text} | Get Text From Document| reference.pdf | #Gets Text Content from .pdf |
| ${text} | Get Text From Document| reference.jpg | #Gets Text Content from .jpg |
| List Should Contain Value | ${text} | Test String | #Checks if list contains a specific string |
"""
img = CompareImage(image)
if img.extension == '.pdf':
text = []
for i in range(len(img.opencv_images)):
tdict = json.loads(img.mupdfdoc[i].get_text("json"))
for block in tdict['blocks']:
if block['type'] == 0:
for line in block['lines']:
if line['spans'][0]['text']:
text.append(line['spans'][0]['text'])
else:
try:
img.get_ocr_text_data()
text = [x for x in img.text_content[0]['text'] if x]
except:
text = None
return text
def test_multipage_pdf(testdata_dir):
img = CompareImage(testdata_dir / 'sample.pdf')
assert len(img.opencv_images) == 2
assert type(img.opencv_images) == list
assert type(img.opencv_images[0]) == numpy.ndarray
assert type(img.opencv_images[1]) == numpy.ndarray
pass
def test_pdf_text_content(testdata_dir):
img = CompareImage(testdata_dir / 'sample_1_page.pdf')
assert len(img.mupdfdoc.get_page_text(0, "WORDS")) > 0
def test_image_text_content(testdata_dir):
img = CompareImage(testdata_dir / 'Beach_date.png')
img.get_ocr_text_data()
assert "01-Jan-2021" in img.text_content[0]['text']
assert "123456789" in img.text_content[0]['text']
def test_single_pdf(testdata_dir):
img = CompareImage(testdata_dir / 'sample_1_page.pdf')
assert len(img.opencv_images) == 1
assert type(img.opencv_images) == list
assert type(img.opencv_images[0]) == numpy.ndarray
pass
def test_image_area_mask(testdata_dir):
img = CompareImage(testdata_dir / 'Beach_date.png',
placeholder_file=testdata_dir / 'area_mask.json')
assert len(img.placeholders) == 1
img_with_mask = img.get_image_with_placeholders()
assert img != img_with_mask
def test_simple_text_from_pdf(testdata_dir):
img = CompareImage(testdata_dir / 'sample_1_page.pdf')
img.get_ocr_text_data()
assert 'FB1DES0A3D5EFE2A60B0B1AE616C653' in img.text_content[0]['text']
def test_single_png(testdata_dir):
img = CompareImage(testdata_dir / 'text_big.png')
assert len(img.opencv_images) == 1
assert type(img.opencv_images) == list
assert type(img.opencv_images[0]) == numpy.ndarray
def test_single_png_with_barcode(testdata_dir):
img = CompareImage(testdata_dir / 'datamatrix.png', contains_barcodes=True)
assert len(img.placeholders) == 2
def test_single_pdf_without_barcode(testdata_dir):
img = CompareImage(testdata_dir / 'sample_1_page.pdf',
contains_barcodes=True)
assert len(img.placeholders) == 0
def check_for_differences(self, reference, candidate, i, detected_differences, compare_options, reference_pdf_content=None, candidate_pdf_content=None):
images_are_equal = True
with futures.ThreadPoolExecutor(max_workers=2) as parallel_executor:
grayA_future = parallel_executor.submit(cv2.cvtColor, reference, cv2.COLOR_BGR2GRAY)
grayB_future = parallel_executor.submit(cv2.cvtColor, candidate, cv2.COLOR_BGR2GRAY)
grayA = grayA_future.result()
grayB = grayB_future.result()
if reference.shape[0] != candidate.shape[0] or reference.shape[1] != candidate.shape[1]:
self.add_screenshot_to_log(reference, "_reference_page_" + str(i+1))
self.add_screenshot_to_log(candidate, "_candidate_page_" + str(i+1))
raise AssertionError(f'The compared images have different dimensions:\nreference:{reference.shape}\ncandidate:{candidate.shape}')
# compute the Structural Similarity Index (SSIM) between the two
# images, ensuring that the difference image is returned
(score, diff) = metrics.structural_similarity(grayA, grayB, gaussian_weights=True, full=True)
score = abs(1-score)
if self.take_screenshots:
# Not necessary to take screenshots for every successful comparison
self.add_screenshot_to_log(np.concatenate((reference, candidate), axis=1), "_page_" + str(i+1) + "_compare_concat")
if (score > self.threshold):
diff = (diff * 255).astype("uint8")
thresh = cv2.threshold(diff, 0, 255,
cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)[1]
reference_with_rect, candidate_with_rect , cnts= self.get_images_with_highlighted_differences(thresh, reference.copy(), candidate.copy(), extension=int(os.getenv('EXTENSION', 2)))
blended_images = self.overlay_two_images(reference_with_rect, candidate_with_rect)
cv2.putText(reference_with_rect,self.REFERENCE_LABEL, self.BOTTOM_LEFT_CORNER_OF_TEXT, self.FONT, self.FONT_SCALE, self.FONT_COLOR, self.LINE_TYPE)
cv2.putText(candidate_with_rect,self.CANDIDATE_LABEL, self.BOTTOM_LEFT_CORNER_OF_TEXT, self.FONT, self.FONT_SCALE, self.FONT_COLOR, self.LINE_TYPE)
self.add_screenshot_to_log(np.concatenate((reference_with_rect, candidate_with_rect), axis=1), "_page_" + str(i+1) + "_rectangles_concat")
self.add_screenshot_to_log(blended_images, "_page_" + str(i+1) + "_blended")
if self.show_diff:
self.add_screenshot_to_log(np.concatenate((diff, thresh), axis=1), "_page_" + str(i+1) + "_diff")
images_are_equal=False
if (compare_options["ignore_watermarks"] == True and len(cnts)==1) or compare_options["watermark_file"] is not None:
if (compare_options["ignore_watermarks"] == True and len(cnts)==1):
(x, y, w, h) = cv2.boundingRect(cnts[0])
diff_center_x = abs((x+w/2)-(reference.shape[1]/2))
diff_center_y = abs((y+h/2)-(reference.shape[0]/2))
if (diff_center_x < reference.shape[1] * self.WATERMARK_CENTER_OFFSET) and (w * 25.4 / self.DPI < self.WATERMARK_WIDTH) and (h * 25.4 / self.DPI < self.WATERMARK_HEIGHT):
images_are_equal=True
print("A watermark position was identified. After ignoring watermark area, both images are equal")
return
if compare_options["watermark_file"] is not None:
watermark_file = compare_options["watermark_file"]
if isinstance(watermark_file, str):
if os.path.isdir(watermark_file):
watermark_file = [str(os.path.join(watermark_file, f)) for f in os.listdir(watermark_file) if os.path.isfile(os.path.join(watermark_file, f))]
else:
watermark_file = [watermark_file]
if isinstance(watermark_file, list):
try:
for single_watermark in watermark_file:
try:
watermark = CompareImage(single_watermark, DPI=self.DPI).opencv_images[0]
except:
print(f'Watermark file {single_watermark} could not be loaded. Continue with next item.')
continue
watermark_gray = cv2.cvtColor(watermark, cv2.COLOR_BGR2GRAY)
watermark_gray = (watermark_gray * 255).astype("uint8")
mask = cv2.threshold(watermark_gray, 10, 255, cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)[1]
mask = cv2.dilate(mask, None, iterations=1)
mask_inv = cv2.bitwise_not(mask)
if thresh.shape[0:2] == mask_inv.shape[0:2]:
result = cv2.bitwise_and(thresh, thresh, mask=mask_inv)
else:
print(f"The shape of watermark and image are different. Continue with next item")
print(f"Document: {thresh.shape}\nMask: {mask_inv.shape}")
continue
if self.show_diff:
print(f"The diff after watermark removal")
self.add_screenshot_to_log(result, "_page_" + str(i + 1) + "_watermark_diff")
if cv2.countNonZero(result) == 0:
images_are_equal=True
print("A watermark file was provided. After removing watermark area, both images are equal")
return
except:
raise AssertionError('The provided watermark_file format is invalid. Please provide a path to a file or a list of files.')
else:
raise AssertionError('The provided watermark_file format is invalid. Please provide a path to a file or a list of files.')
if(compare_options["check_text_content"]==True) and images_are_equal is not True:
if compare_options["get_pdf_content"] is not True:
#x, y, w, h = self.get_diff_rectangle(thresh)
images_are_equal=True
for c in range(len(cnts)):
(x, y, w, h) = cv2.boundingRect(cnts[c])
diff_area_reference = reference[y:y+h, x:x+w]
diff_area_candidate = candidate[y:y+h, x:x+w]
self.add_screenshot_to_log(diff_area_reference, "_page_" + str(i+1) + "_diff_area_reference_"+str(c))
self.add_screenshot_to_log(diff_area_candidate, "_page_" + str(i+1) + "_diff_area_test_"+str(c))
text_reference = pytesseract.image_to_string(diff_area_reference, config='--psm 6').replace("\n\n", "\n")
text_candidate = pytesseract.image_to_string(diff_area_candidate, config='--psm 6').replace("\n\n", "\n")
if text_reference.strip()==text_candidate.strip():
print("Partial text content is the same")
print(text_reference)
else:
images_are_equal=False
detected_differences.append(True)
print("Partial text content is different")
print(text_reference + " is not equal to " + text_candidate)
elif compare_options["get_pdf_content"] is True:
images_are_equal=True
ref_words = reference_pdf_content.get_text("words")
cand_words = candidate_pdf_content.get_text("words")
for c in range(len(cnts)):
(x, y, w, h) = cv2.boundingRect(cnts[c])
rect = fitz.Rect(x*72/self.DPI, y*72/self.DPI, (x+w)*72/self.DPI, (y+h)*72/self.DPI)
diff_area_ref_words = [w for w in ref_words if fitz.Rect(w[:4]).intersects(rect)]
diff_area_cand_words = [w for w in cand_words if fitz.Rect(w[:4]).intersects(rect)]
diff_area_ref_words = make_text(diff_area_ref_words)
diff_area_cand_words = make_text(diff_area_cand_words)
diff_area_reference = reference[y:y+h, x:x+w]
diff_area_candidate = candidate[y:y+h, x:x+w]
self.add_screenshot_to_log(diff_area_reference, "_page_" + str(i+1) + "_diff_area_reference_"+str(c))
self.add_screenshot_to_log(diff_area_candidate, "_page_" + str(i+1) + "_diff_area_test_"+str(c))
if len(diff_area_ref_words)!=len(diff_area_cand_words):
images_are_equal=False
detected_differences.append(True)
print("The identified pdf layout elements are different", diff_area_ref_words, diff_area_cand_words)
else:
if diff_area_ref_words.strip() != diff_area_cand_words.strip():
images_are_equal=False
detected_differences.append(True)
print("Partial text content is different")
print(diff_area_ref_words.strip(), " is not equal to " ,diff_area_cand_words.strip())
if images_are_equal:
print("Partial text content of area is the same")
print(diff_area_ref_words)
pass
if(compare_options["move_tolerance"]!=None) and images_are_equal is not True:
move_tolerance=int(compare_options["move_tolerance"])
images_are_equal=True
if compare_options["get_pdf_content"] is not True:
#Experimental, to solve a problem with small images
#wr, hr, _ = reference.shape
for c in range(len(cnts)):
(x, y, w, h) = cv2.boundingRect(cnts[c])
diff_area_reference = reference[y:y+h, x:x+w]
diff_area_candidate = candidate[y:y+h, x:x+w]
#Experimental, to solve a problem with small images
#search_area_candidate = candidate[(y - self.BORDER_FOR_MOVE_TOLERANCE_CHECK) if y >= self.BORDER_FOR_MOVE_TOLERANCE_CHECK else 0:(y + h + self.BORDER_FOR_MOVE_TOLERANCE_CHECK) if hr >= (y + h + self.BORDER_FOR_MOVE_TOLERANCE_CHECK) else hr, (x - self.BORDER_FOR_MOVE_TOLERANCE_CHECK) if x >= self.BORDER_FOR_MOVE_TOLERANCE_CHECK else 0:(x + w + self.BORDER_FOR_MOVE_TOLERANCE_CHECK) if wr >= (x + w + self.BORDER_FOR_MOVE_TOLERANCE_CHECK) else wr]
search_area_candidate = candidate[y - self.BORDER_FOR_MOVE_TOLERANCE_CHECK:y + h + self.BORDER_FOR_MOVE_TOLERANCE_CHECK, x - self.BORDER_FOR_MOVE_TOLERANCE_CHECK:x + w + self.BORDER_FOR_MOVE_TOLERANCE_CHECK]
search_area_reference = reference[y - self.BORDER_FOR_MOVE_TOLERANCE_CHECK:y + h + self.BORDER_FOR_MOVE_TOLERANCE_CHECK, x - self.BORDER_FOR_MOVE_TOLERANCE_CHECK:x + w + self.BORDER_FOR_MOVE_TOLERANCE_CHECK]
# self.add_screenshot_to_log(search_area_candidate)
# self.add_screenshot_to_log(search_area_reference)
# self.add_screenshot_to_log(diff_area_candidate)
# self.add_screenshot_to_log(diff_area_reference)
try:
positions_in_compare_image = self.find_partial_image_position(search_area_candidate, diff_area_reference)
except:
print("Error in finding position in compare image")
images_are_equal=False
detected_differences.append(True)
continue
#positions_in_compare_image = self.find_partial_image_position(candidate, diff_area_reference)
if (np.mean(diff_area_reference) == 255) or (np.mean(diff_area_candidate) == 255):
images_are_equal=False
detected_differences.append(True)
print("Image section contains only white background")
self.add_screenshot_to_log(np.concatenate((cv2.copyMakeBorder(diff_area_reference, top=2, bottom=2, left=2, right=2, borderType=cv2.BORDER_CONSTANT, value=[0,0,0]), cv2.copyMakeBorder(diff_area_candidate, top=2, bottom=2, left=2, right=2, borderType=cv2.BORDER_CONSTANT, value=[0,0,0])), axis=1), "_diff_area_concat")
#self.add_screenshot_to_log(np.concatenate((diff_area_reference, diff_area_candidate), axis=1), "_diff_area_concat")
else:
if positions_in_compare_image:
#pt_original = (x, y)
pt_original = positions_in_compare_image['pt1']
pt_compare = positions_in_compare_image['pt2']
x_moved = abs(pt_original[0]-pt_compare[0])
y_moved = abs(pt_original[1]-pt_compare[1])
move_distance = math.sqrt(x_moved** 2 +y_moved ** 2)
#cv2.arrowedLine(candidate_with_rect, pt_original, pt_compare, (255, 0, 0), 4)
if int(move_distance)>int(move_tolerance):
print("Image section moved ",move_distance, " pixels")
print("This is outside of the allowed range of ",move_tolerance, " pixels")
images_are_equal=False
detected_differences.append(True)
self.add_screenshot_to_log(self.overlay_two_images(search_area_reference, search_area_candidate), "_diff_area_blended")
else:
print("Image section moved ",move_distance, " pixels")
print("This is within the allowed range of ",move_tolerance, " pixels")
self.add_screenshot_to_log(self.overlay_two_images(search_area_reference, search_area_candidate), "_diff_area_blended")
else:
images_are_equal=False
detected_differences.append(True)
print("The reference image section was not found in test image (or vice versa)")
self.add_screenshot_to_log(np.concatenate((cv2.copyMakeBorder(diff_area_reference, top=2, bottom=2, left=2, right=2, borderType=cv2.BORDER_CONSTANT, value=[0,0,0]), cv2.copyMakeBorder(diff_area_candidate, top=2, bottom=2, left=2, right=2, borderType=cv2.BORDER_CONSTANT, value=[0,0,0])), axis=1), "_diff_area_concat")
elif compare_options["get_pdf_content"] is True:
images_are_equal=True
ref_words = reference_pdf_content.get_text("words")
cand_words = candidate_pdf_content.get_text("words")
for c in range(len(cnts)):
(x, y, w, h) = cv2.boundingRect(cnts[c])
rect = fitz.Rect(x*72/self.DPI, y*72/self.DPI, (x+w)*72/self.DPI, (y+h)*72/self.DPI)
diff_area_ref_words = [w for w in ref_words if fitz.Rect(w[:4]).intersects(rect)]
diff_area_cand_words = [w for w in cand_words if fitz.Rect(w[:4]).intersects(rect)]
# diff_area_ref_words = make_text(diff_area_ref_words)
# diff_area_cand_words = make_text(diff_area_cand_words)
diff_area_reference = reference[y:y+h, x:x+w]
diff_area_candidate = candidate[y:y+h, x:x+w]
self.add_screenshot_to_log(diff_area_reference, "_page_" + str(i+1) + "_diff_area_reference_"+str(c))
self.add_screenshot_to_log(diff_area_candidate, "_page_" + str(i+1) + "_diff_area_test_"+str(c))
if len(diff_area_ref_words)!=len(diff_area_cand_words):
images_are_equal=False
detected_differences.append(True)
print("The identified pdf layout elements are different", diff_area_ref_words, diff_area_cand_words)
else:
for ref_Item, cand_Item in zip(diff_area_ref_words, diff_area_cand_words):
if ref_Item == cand_Item:
pass
elif str(ref_Item[4]).strip() == str(cand_Item[4]).strip():
left_moved = abs(ref_Item[0]-cand_Item[0])*self.DPI/72
top_moved = abs(ref_Item[1]-cand_Item[1])*self.DPI/72
right_moved = abs(ref_Item[2]-cand_Item[2])*self.DPI/72
bottom_moved = abs(ref_Item[3]-cand_Item[3])*self.DPI/72
print("Checking pdf elements", ref_Item, cand_Item)
if int(left_moved)>int(move_tolerance) or int(top_moved)>int(move_tolerance) or int(right_moved)>int(move_tolerance) or int(bottom_moved)>int(move_tolerance):
print("Image section moved ",left_moved, top_moved, right_moved, bottom_moved, " pixels")
print("This is outside of the allowed range of ",move_tolerance, " pixels")
images_are_equal=False
detected_differences.append(True)
self.add_screenshot_to_log(self.overlay_two_images(diff_area_reference, diff_area_candidate), "_diff_area_blended")
else:
print("Image section moved ",left_moved, top_moved, right_moved, bottom_moved, " pixels")
print("This is within the allowed range of ",move_tolerance, " pixels")
self.add_screenshot_to_log(self.overlay_two_images(diff_area_reference, diff_area_candidate), "_diff_area_blended")
if images_are_equal is not True:
detected_differences.append(True)
def test_big_text_from_image(testdata_dir):
img = CompareImage(testdata_dir / 'text_big.png')
img.get_ocr_text_data()
assert 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' in img.text_content[0]['text']
pass
def test_white_text_on_dark_background(testdata_dir):
img = CompareImage(testdata_dir / 'whitetext_blackbackground.png')
img.get_ocr_text_data()
assert '0123456789' in img.text_content[0]['text']
pass
def test_text_on_colored_background(testdata_dir):
img = CompareImage(testdata_dir / 'Beach_date.png')
img.get_ocr_text_data()
assert "01-Jan-2021" in img.text_content[0]['text']
assert "123456789" in img.text_content[0]['text']
pass
def test_non_existing_file(testdata_dir):
with pytest.raises(AssertionError):
img = CompareImage(testdata_dir / 'does_not_exist.png')
def test_corrupt_pdf(testdata_dir):
with pytest.raises(AssertionError):
img = CompareImage(testdata_dir / 'corrupt_pdf.pdf')
def test_pdf_text_mask(testdata_dir):
img = CompareImage(testdata_dir / 'sample_1_page.pdf',
placeholder_file=testdata_dir / 'pdf_pattern_mask.json')
assert len(img.placeholders) == 3
img_with_mask = img.get_image_with_placeholders()
assert img != img_with_mask
def test_image_text_content_with_east(testdata_dir):
img = CompareImage(testdata_dir / 'Beach_date.png')
img.get_text_content_with_east()
assert any('01-Jan-2021' in s for s in img.text_content[0]['text'])
def test_small_text_from_image(testdata_dir):
img = CompareImage(testdata_dir / 'text_small.png')
img.get_ocr_text_data()
assert '1234567890' in img.text_content[0]['text']
pass