detect an initial/a sketch drawing on a text page - python

I would like to get the coordinates of the box around the initial ("H") on the following page (and similar ones with other initials, so opencv template matching is not an option):
Following this tutorial, I tried to solve the problem with opencv contours:
import cv2
import matplotlib.pyplot as plt
page = "image.jpg"
# read the image
image = cv2.imread(page)
# convert to RGB
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
# convert to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
# create a binary thresholded image
_, binary = cv2.threshold(gray, 0,150,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
# find the contours from the thresholded image
contours, hierarchy = cv2.findContours(binary, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
# draw all contours
image = cv2.drawContours(image, contours, 3, (0, 255, 0), 2)
plt.savefig("result.png")
The result is of course not exactly what I wanted:
Does anyone know of an viable algorithm (and possibly an implementation thereof) that could provide an easy solution to my task?

You can find the target area by filtering your contours. Now, there's at least two filtering criteria that you can use. One is filter by area - that is, discard too small and too large contours until you get the contour you are looking for. The other one is by computing the extent of every contour. The extent is the ratio of the contour's area to its bounding rectangle area. You are looking for a square-like contour, so its extent should be close to 1.0.
Let's see the code:
# imports:
import cv2
import numpy as np
# Reading an image in default mode:
inputImage = cv2.imread(path + fileName)
# Deep copy for results:
inputImageCopy = inputImage.copy()
# Convert RGB to grayscale:
grayscaleImage = cv2.cvtColor(inputImage, cv2.COLOR_BGR2GRAY)
# Get binary image via Otsu:
_, binaryImage = cv2.threshold(grayscaleImage, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
The first portion of the code gets you a binary image that you can use as a mask to compute contours:
Now, let's filter contours. Let's use the area approach first. You need to define a range of minimum area and maximum area to filter everything that does not fall in this range. I've heuristically determined a range of areas from 30000 px to 150000 px:
# Find the contours on the binary image:
contours, hierarchy = cv2.findContours(binaryImage, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# Look for the outer bounding boxes (no children):
for _, c in enumerate(contours):
# Get blob area:
currentArea = cv2.contourArea(c)
print("Contour Area: "+str(currentArea))
# Set an area range:
minArea = 30000
maxArea = 150000
if minArea < currentArea < maxArea:
# Get the contour's bounding rectangle:
boundRect = cv2.boundingRect(c)
# Get the dimensions of the bounding rect:
rectX = boundRect[0]
rectY = boundRect[1]
rectWidth = boundRect[2]
rectHeight = boundRect[3]
# Set bounding rect:
color = (0, 0, 255)
cv2.rectangle( inputImageCopy, (int(rectX), int(rectY)),
(int(rectX + rectWidth), int(rectY + rectHeight)), color, 2 )
cv2.imshow("Rectangles", inputImageCopy)
cv2.waitKey(0)
Once you successfully filter the area, you can then compute the bounding rectangle of the contour with cv2.boundingRect. You can retrieve the bounding rectangle's x, y (top left) coordinates as well as its width and height. After that just draw the rectangle on a deep copy of the original input.
Now, let's see the second option, using the contour's extent. The for loop gets modified as follows:
# Look for the outer bounding boxes (no children):
for _, c in enumerate(contours):
# Get blob area:
currentArea = cv2.contourArea(c)
# Get the contour's bounding rectangle:
boundRect = cv2.boundingRect(c)
# Get the dimensions of the bounding rect:
rectX = boundRect[0]
rectY = boundRect[1]
rectWidth = boundRect[2]
rectHeight = boundRect[3]
# Calculate extent:
extent = float(currentArea)/(rectWidth *rectHeight)
print("Extent: " + str(extent))
# Set the extent filter, look for an extent close to 1.0:
delta = abs(1.0 - extent)
epsilon = 0.1
if delta < epsilon:
# Set bounding rect:
color = (0, 0, 255)
cv2.rectangle( inputImageCopy, (int(rectX), int(rectY)),
(int(rectX + rectWidth), int(rectY + rectHeight)), color, 2 )
cv2.imshow("Rectangles", inputImageCopy)
cv2.waitKey(0)
Both approaches yield this result:

You almost have it. You just need to filter contours on area and aspect ratio. Here is my approach in Python/OpenCV.
Input:
import cv2
import numpy as np
# read image as grayscale
img = cv2.imread('syriados.jpg')
# convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# threshold to binary
#thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY)[1]
thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU)[1]
# invert threshold
thresh = 255 - thresh
# apply morphology to remove small white regions and to close the rectangle boundary
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3,3))
morph = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel)
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (7,7))
morph = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel)
# find contours
result = img.copy()
cntrs = cv2.findContours(morph, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
cntrs = cntrs[0] if len(cntrs) == 2 else cntrs[1]
# filter on area and aspect ratio
for c in cntrs:
area = cv2.contourArea(c)
x,y,w,h = cv2.boundingRect(c)
if area > 10000 and abs(w-h) < 100:
cv2.drawContours(result, [c], 0, (0,0,255), 2)
# write results
cv2.imwrite("syriados_thresh.jpg", thresh)
cv2.imwrite("syriados_morph.jpg", morph)
cv2.imwrite("syriados_box.jpg", result)
# show results
cv2.imshow("thresh", thresh)
cv2.imshow("morph", morph)
cv2.imshow("result", result)
cv2.waitKey(0)
Threshold image:
Morphology image:
Resulting contour image:

To get a result like this:
You'll need to detect the contour in the image with the second to the greatest area, as the one possessing the greatest area would be the border of the image.
So with the list of contours, we can get the one with the second greatest area via the built-in sorted method, using the cv2.contourArea method as the custom key:
import cv2
import numpy as np
def process(img):
img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img_blur = cv2.GaussianBlur(img_gray, (7, 7), 2)
img_canny = cv2.Canny(img_blur, 50, 50)
kernel = np.ones((6, 6))
img_dilate = cv2.dilate(img_canny, kernel, iterations=1)
img_erode = cv2.erode(img_dilate, kernel, iterations=2)
return img_erode
def get_contours(img):
contours, _ = cv2.findContours(process(img), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
cnt = sorted(contours, key=cv2.contourArea)[-2]
peri = cv2.arcLength(cnt, True)
approx = cv2.approxPolyDP(cnt, 0.02 * peri, True)
cv2.drawContours(img, [approx], -1, (0, 255, 0), 2)
page = "image.jpg"
image = cv2.imread(page)
get_contours(image)
cv2.imshow("Image", image)
cv2.waitKey(0)
The above only puts the area of the contours into consideration; if you want more reliable results, you can make it so that it will only detect contours that are 4-sided.

Related

Detect thick black lines in image with OpenCV

I have the following image of a lego board with some bricks on it
Now I am trying to detect the thick black lines (connecting the white squares) with OpenCV. I have already experimented a lot with HoughLinesP, converted the image to gray or b/w before, applied blur, ...
Nonthing led to usable results.
# Read image
img = cv2.imread('image.jpg', cv2.IMREAD_GRAYSCALE)
# Resize Image
img = cv2.resize(img, (0,0), fx=0.25, fy=0.25)
# Initialize output
out = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
# Median blurring to get rid of the noise; invert image
img = cv2.medianBlur(img, 5)
# Adaptive Treshold
bw = cv2.adaptiveThreshold(img,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,\
cv2.THRESH_BINARY,15,8)
# HoughLinesP
linesP = cv2.HoughLinesP(bw, 500, np.pi / 180, 50, None, 50, 10)
# Draw Lines
if linesP is not None:
for i in range(0, len(linesP)):
l = linesP[i][0]
cv2.line(out, (l[0], l[1]), (l[2], l[3]), (0,0,255), 3, cv2.LINE_AA)
The adaptive treshold lets you see edges quite well, but with HoughLinesP you don't get anything usable out of it
What am I doing wrong?
Thanks, both #fmw42 and #jeru-luke for your great solutions to this problem! I liked isolating / masking the green board, so I combined both:
import cv2
import numpy as np
img = cv2.imread("image.jpg")
scale_percent = 50 # percent of original size
width = int(img.shape[1] * scale_percent / 100)
height = int(img.shape[0] * scale_percent / 100)
dim = (width, height)
# resize image
img = cv2.resize(img, dim, interpolation = cv2.INTER_AREA)
lab = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
a_component = lab[:,:,1]
# binary threshold the a-channel
th = cv2.threshold(a_component,127,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)[1]
# numpy black
black = np.zeros((img.shape[0],img.shape[1]),np.uint8)
# function to obtain the largest contour in given image after filling it
def get_region(image):
contours, hierarchy = cv2.findContours(image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
c = max(contours, key = cv2.contourArea)
mask = cv2.drawContours(black,[c],0,255, -1)
return mask
mask = get_region(th)
# turning the region outside the green block white
green_block = cv2.bitwise_and(img, img, mask = mask)
green_block[black==0]=(255,255,255)
# median blur
median = cv2.medianBlur(green_block, 5)
# threshold on black
lower = (0,0,0)
upper = (15,15,15)
thresh = cv2.inRange(median, lower, upper)
# apply morphology open and close
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3,3))
morph = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel)
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (29,29))
morph = cv2.morphologyEx(morph, cv2.MORPH_CLOSE, kernel)
# filter contours on area
contours = cv2.findContours(morph, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
contours = contours[0] if len(contours) == 2 else contours[1]
result = green_block.copy()
for c in contours:
area = cv2.contourArea(c)
if area > 1000:
cv2.drawContours(result, [c], -1, (0, 0, 255), 2)
# view result
cv2.imshow("result", result)
cv2.waitKey(0)
cv2.destroyAllWindows()
Here I am presenting a repeated segmentation approach using color.
This answer is based on the usage of LAB color space
1. Isolating the green lego block
img = cv2.imread(image_path)
lab = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
a_component = lab[:,:,1]
# binary threshold the a-channel
th = cv2.threshold(a_component,127,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)[1]
th
# function to obtain the largest contour in given image after filling it
def get_region(image):
contours, hierarchy = cv2.findContours(image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
c = max(contours, key = cv2.contourArea)
black = np.zeros((image.shape[0], image.shape[1]), np.uint8)
mask = cv2.drawContours(black,[c],0,255, -1)
return mask
mask = get_region(th)
mask
# turning the region outside the green block white
green_block = cv2.bitwise_and(img, img, mask = mask)
green_block[black==0]=(255,255,255)
green_block
2. Segmenting the road
To get an approximate region of the road, I subtracted the mask and th.
cv2.subtract() performs arithmetic subtraction, where cv2 will take care of negative values.
road = cv2.subtract(mask,th)
# `road` contains some unwanted spots/contours which are removed using the function "get_region"
only_road = get_region(road)
only_road
Masking only the road segment with the original image gives
road_colored = cv2.bitwise_and(img, img, mask = only_road)
road_colored[only_road==0]=(255,255,255)
road_colored
From the above image only the black regions (road) are present, which is easy to segment:
# converting to grayscale and applying threshold
th2 = cv2.threshold(road_colored[:,:,1],127,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)[1]
# using portion of the code from fmw42's answer, to get contours above certain area
contours = cv2.findContours(th2, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
contours = contours[0] if len(contours) == 2 else contours[1]
result = img.copy()
for c in contours:
area = cv2.contourArea(c)
if area > 1000:
cv2.drawContours(result, [c], -1, (0, 0, 255), 4)
result
Note:
To clean up the end result, you can apply morphological operations on th2 before drawing contours.
Here is one way to do that in Python/OpenCV.
Read the image
Apply median blur
Threshold on black color using cv2.inRange()
Apply morphology to clean it up
Get contours and filter on area
Draw contours on input
Save the result
Input:
import cv2
import numpy as np
# read image
img = cv2.imread('black_lines.jpg')
# median blur
median = cv2.medianBlur(img, 5)
# threshold on black
lower = (0,0,0)
upper = (15,15,15)
thresh = cv2.inRange(median, lower, upper)
# apply morphology open and close
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3,3))
morph = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel)
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (29,29))
morph = cv2.morphologyEx(morph, cv2.MORPH_CLOSE, kernel)
# filter contours on area
contours = cv2.findContours(morph, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
contours = contours[0] if len(contours) == 2 else contours[1]
result = img.copy()
for c in contours:
area = cv2.contourArea(c)
if area > 1000:
cv2.drawContours(result, [c], -1, (0, 0, 255), 2)
# save result
cv2.imwrite("black_lines_threshold.jpg", thresh)
cv2.imwrite("black_lines_morphology.jpg", morph)
cv2.imwrite("black_lines_result.jpg", result)
# view result
cv2.imshow("threshold", thresh)
cv2.imshow("morphology", morph)
cv2.imshow("result", result)
cv2.waitKey(0)
cv2.destroyAllWindows()
Threshold image:
Morphology image:
Result:

Python - cv2 find Contours

I would like to find all the big elements in the document, but I do not know how to control the size (the document is downloaded from the Internet :))
I have a document
And I wrote a simple code
import cv2
import pytesseract
image = cv2.imread('2.png')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (7, 7), 0)
thresh = cv2.threshold(
blur, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
kernal = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 50))
dilate = cv2.dilate(thresh, kernal, iterations=1)
cv2.imwrite('1_dilated.png', dilate)
cnts = cv2.findContours(dilate, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]
cnts = sorted(cnts, key=lambda x: cv2.boundingRect(x)[1])
for c in cnts:
x, y, w, h = cv2.boundingRect(c)
if h > 100 and w > 100:
roi = image[y:y+h, x:x+w]
cv2.rectangle(image, (x, y), (x+w, y+h), (36, 255, 12), 2)
# ocr = pytesseract.image_to_string(roi)
# print(ocr)
cv2.imwrite('1_boxes4.png', image)
But only detects it
And I would like this
How to control the size of the detected area ?
Thank you very much for all your comments
You are close, but you need to increase the number of iterations of the dilate operation. Also, a rectangular structuring element might help better forming the blobs of text. Let's check out some possible improvements of your code:
# imports:
import cv2
import numpy as np
# Set image path
imagePath = "D://opencvImages//"
imageName = "F74Yq.png"
# Read image:
inputImage = cv2.imread(imagePath + imageName)
# Store a deeep copy for results:
inputCopy = inputImage.copy()
# Convert BGR to grayscale:
grayInput = cv2.cvtColor(inputImage, cv2.COLOR_BGR2GRAY)
# Threshold via Otsu
_, binaryImage = cv2.threshold(grayInput, 0, 255, cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
The first part produces the binary image of the input image, there's nothing fancy going on here - just a direct thresholding via Otsu's method. This is the binary image obtained:
Now, let's apply the dilate operation. Let's use a 9 x 9 rectangular kernel and set the number of iterations to 5. Gotta be careful you don't dilate too much, because blobs of text from different portions of the document could end up joined:
# Set kernel (structuring element) size:
kernelSize = (9, 9)
# Set operation iterations:
opIterations = 5
# Get the structuring element:
morphKernel = cv2.getStructuringElement(cv2.MORPH_RECT, kernelSize)
# Perform Dilate:
dilateImage = cv2.morphologyEx(binaryImage, cv2.MORPH_DILATE, morphKernel, None, None, opIterations, cv2.BORDER_REFLECT101)
This is the result:
Ok, now let's just detect external contours and get their bounding boxes so we can draw rectangles around the target areas. Note that I'm drawing the rectangles on a deep copy of the input:
# Find the contours on the binary image:
contours, hierarchy = cv2.findContours(dilateImage, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# Look for the outer bounding boxes (no children):
for _, c in enumerate(contours):
# Get the contours bounding rectangle:
boundRect = cv2.boundingRect(c)
# Get the dimensions of the bounding rectangle:
rectX = boundRect[0]
rectY = boundRect[1]
rectWidth = boundRect[2]
rectHeight = boundRect[3]
# Set bounding rectangle:
color = (0, 0, 255)
cv2.rectangle( inputCopy, (int(rectX), int(rectY)),
(int(rectX + rectWidth), int(rectY + rectHeight)), color, 5 )
cv2.imshow("Bounding Rectangles", inputCopy)
cv2.waitKey()
This is the final result:

How to remove text from the sketched image

I have some sketched images where the images contain text captions. I am trying to remove those caption.
I am using this code:
import cv2
import pytesseract
pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe"
# Load image, grayscale, blur, Otsu's threshold
image = cv2.imread('1.png')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (3,3), 0)
thresh = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
# Find contours and filter using contour area
cnts = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]
for c in cnts:
area = cv2.contourArea(c)
if area > 500:
cv2.drawContours(thresh, [c], -1, 0, -1)
# Invert image and OCR
invert = 255 - thresh
Output= thresh - invert
cv2.imshow('thresh', thresh)
cv2.imshow('invert', invert)
cv2.imshow('output', output)
cv2.waitKey()
Code not working for these images.
The cv2 pre-processing is unecessary here, tesseract is able to find the text on its own. See the example below, commented inline:
results = pytesseract.image_to_data('1.png', config='--psm 11', output_type='dict')
for i in range(len(results["text"])):
# extract the bounding box coordinates of the text region from
# the current result
x = results["left"][i]
y = results["top"][i]
w = results["width"][i]
h = results["height"][i]
# Extract the confidence of the text
conf = int(results["conf"][i])
if conf > 60: # adjust to your liking
# Cover the text with a white rectangle
cv2.rectangle(image, (x, y), (x + w, y + h), (255, 255, 255), -1)
Detected text on the left, cleaned image on the right:
Another option, without using Tesseract. Just use the area of the contours to filter the smaller ones by covering them with white-filled rectangles:
# Imports
import cv2
import numpy as np
# Read image
imagePath = "C://opencvImages//"
inputImage = cv2.imread(imagePath+"0enxN.png")
# Convert BGR to grayscale:
binaryImage = cv2.cvtColor(inputImage, cv2.COLOR_BGR2GRAY)
# Invert image:
binaryImage = 255 - binaryImage
# Find the external contours on the binary image:
contours, hierarchy = cv2.findContours(binaryImage, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# Invert image:
binaryImage = 255 - binaryImage
# Look for the bounding boxes:
for _, c in enumerate(contours):
# Get the contour's bounding rectangle:
boundRect = cv2.boundingRect(c)
# Get the dimensions of the bounding rect:
rectX = boundRect[0]
rectY = boundRect[1]
rectWidth = boundRect[2]
rectHeight = boundRect[3]
# Get Bounding Rectangle Area:
rectArea = rectWidth * rectHeight
# Set minimum area threshold:
minArea = 1000
# Check for minimum area:
if rectArea < minArea:
# Draw white rectangle to cover small contour:
cv2.rectangle(binaryImage, (rectX, rectY), (rectX + rectWidth, rectY + rectHeight),
(255, 255, 255), -1)
cv2.imshow("Binary Mask", binaryImage)
cv2.waitKey(0)
This produces:

Enhancing corner detection of lamp with OpenCV

I'm using the following code to detect the brightly illuminated lamp. The illumination might vary. I'm using the following code to detect the same.
img = cv2.imread("input_img.jpg")
rgb = img.copy()
img_grey = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
while True:
th3 = cv2.adaptiveThreshold(img_grey, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, \
cv2.THRESH_BINARY, 11, 2)
cv2.imshow("th3",th3)
edged = cv2.Canny(th3, 50, 100)
edged = cv2.dilate(edged, None, iterations=1)
edged = cv2.erode(edged, None, iterations=1)
cv2.imshow("edge", edged)
cnts = cv2.findContours(edged.copy(), cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
cnts = imutils.grab_contours(cnts)
areaArray = []
for i, c in enumerate(cnts):
area = cv2.contourArea(c)
areaArray.append(area)
sorteddata = sorted(zip(areaArray, cnts), key=lambda x: x[0], reverse=True)
thirdlargestcontour = sorteddata[2][1]
x, y, w, h = cv2.boundingRect(thirdlargestcontour)
cv2.drawContours(rgb, thirdlargestcontour, -1, (255, 0, 0), 2)
cv2.rectangle(rgb, (x, y), (x + w, y + h), (0, 255, 0), 2)
cv2.imshow("rgb", rgb)
if cv2.waitKey(1) == 27:
break
The above code works but,
It only gives the rectangle that encompasses the lamp. How do I get the four corner points of the lamp precisely?
How can I improve detection? at the moment I'm picking the third-largest contour which does not guarantee that it will always be the lamp as the environment poses challenge?
ApproxPolydp works when the contour is complete but if the contour is incomplete, ApproxPolydp is not returning the proper coordinate. for instance in the following image the approxpolydp returns a wrong coordinates.
Here is one way to do that in Python/OpenCV.
Read the input image and convert to grayscale
Use adaptive thresholding to get a thick outline of the lamp region
Find the contours
Filter the contours on area to remove extraneous regions and keep only the larger of the two (inner and outer contours of thresholded region)
Get the perimeter
Fit the perimeter to a polygon, which should be a quadrilateral with the right choice of arguments.
Draw the contour (red) and polygon (blue) over a copy of the input image as the result
Input:
import cv2
import numpy as np
# load image
img = cv2.imread("lamp.jpg")
# convert to gray
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# threshold image
thresh = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 11, 10)
thresh = 255 - thresh
# find contours
cntrs = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
cntrs = cntrs[0] if len(cntrs) == 2 else cntrs[1]
# Contour filtering -- remove small objects and those that are too large
# Keep the larger of the two contours (inner and outer contours from thresh)
area_thresh = 0
for c in cntrs:
area = cv2.contourArea(c)
if area > 200 and area > area_thresh:
big_contour = c
area_thresh = area
# draw big_contour on image in red and polygon in blue and print corners
results = img.copy()
cv2.drawContours(results,[big_contour],0,(0,0,255),1)
peri = cv2.arcLength(big_contour, True)
corners = cv2.approxPolyDP(big_contour, 0.04 * peri, True)
cv2.drawContours(results,[corners],0,(255,0,0),1)
print(len(corners))
print(corners)
# write result to disk
cv2.imwrite("lamp_thresh.jpg", thresh)
cv2.imwrite("lamp_corners.jpg", results)
cv2.imshow("THRESH", thresh)
cv2.imshow("RESULTS", results)
cv2.waitKey(0)
cv2.destroyAllWindows()
Thresholded Image:
Result Image:
Corner Coordinates:
[[[233 145]]
[[219 346]]
[[542 348]]
[[508 153]]]

Finding the corners of a rectangle

I'm trying to get the corners of this rectangle:
.
I tried using cv2.cornerHarris(rectangle, 2, 3, 0.04), but the left edges are not showed due to image brightness, I guess. So I tried applying a threshold before using cornerHarris, but the image produced showed a lot of vertices along the edges, not being possible to filter the corners.
I know that I need to filter it before using cornerHarris, but I don't know how. Could someone help me with this problem?
Ps. I've already tried to use blur, but it also doesn't work.
import cv2
import numpy as np
import matplotlib.pyplot as plt
rectangle = cv2.imread('rectangle.png', cv2.IMREAD_GRAYSCALE)
rectangle = np.where(rectangle > np.mean(rectangle), 255, 0).astype(np.uint8)
dst_rectangle = cv2.cornerHarris(rectangle, 2, 3, 0.04)
dst_rectangle = cv2.dilate(dst_rectangle, None)
mask = np.where(dst_rectangle > 0.01*np.max(dst_rectangle), 255, 0).astype(np.uint8)
points = np.nonzero(mask)
plt.imshow(dst_rectangle, cmap='gray')
plt.plot(points[1], points[0], 'or')
plt.show()
I would approach it differently by getting the corners of the rotated bounding box of the contour after adaptive thresholding. Here is my code in Python/OpenCV.
Input:
import cv2
import numpy as np
# read image
img = cv2.imread("rectangle.png")
# convert img to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
gray = 255-gray
# do adaptive threshold on gray image
thresh = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 17, 1)
thresh = 255-thresh
# apply morphology
kernel = np.ones((3,3), np.uint8)
morph = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel)
morph = cv2.morphologyEx(morph, cv2.MORPH_CLOSE, kernel)
# separate horizontal and vertical lines to filter out spots outside the rectangle
kernel = np.ones((7,3), np.uint8)
vert = cv2.morphologyEx(morph, cv2.MORPH_OPEN, kernel)
kernel = np.ones((3,7), np.uint8)
horiz = cv2.morphologyEx(morph, cv2.MORPH_OPEN, kernel)
# combine
rect = cv2.add(horiz,vert)
# thin
kernel = np.ones((3,3), np.uint8)
rect = cv2.morphologyEx(rect, cv2.MORPH_ERODE, kernel)
# get largest contour
contours = cv2.findContours(rect, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
contours = contours[0] if len(contours) == 2 else contours[1]
for c in contours:
area_thresh = 0
area = cv2.contourArea(c)
if area > area_thresh:
area = area_thresh
big_contour = c
# get rotated rectangle from contour
rot_rect = cv2.minAreaRect(big_contour)
box = cv2.boxPoints(rot_rect)
box = np.int0(box)
print(box)
# draw rotated rectangle on copy of img
rot_bbox = img.copy()
cv2.drawContours(rot_bbox,[box],0,(0,0,255),2)
# write img with red rotated bounding box to disk
cv2.imwrite("rectangle_thresh.png", thresh)
cv2.imwrite("rectangle_outline.png", rect)
cv2.imwrite("rectangle_bounds.png", rot_bbox)
# display it
cv2.imshow("IMAGE", img)
cv2.imshow("THRESHOLD", thresh)
cv2.imshow("MORPH", morph)
cv2.imshow("VERT", vert)
cv2.imshow("HORIZ", horiz)
cv2.imshow("RECT", rect)
cv2.imshow("BBOX", rot_bbox)
cv2.waitKey(0)
Thresholded Image:
Rectangle Region Extracted:
Rotated Bounding Box on Image:
Rotated Bounding Box Corners:
[[446 335]
[163 328]
[168 117]
[451 124]]
ADDITION:
Here is a slightly shorter version of the code, which is achievable by adding some gaussian blurring before thresholding.
import cv2
import numpy as np
# read image
img = cv2.imread("rectangle.png")
# convert img to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
gray = 255-gray
# blur image
blur = cv2.GaussianBlur(gray, (3,3), 0)
# do adaptive threshold on gray image
thresh = cv2.adaptiveThreshold(blur, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 75, 2)
thresh = 255-thresh
# apply morphology
kernel = np.ones((5,5), np.uint8)
rect = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel)
rect = cv2.morphologyEx(rect, cv2.MORPH_CLOSE, kernel)
# thin
kernel = np.ones((5,5), np.uint8)
rect = cv2.morphologyEx(rect, cv2.MORPH_ERODE, kernel)
# get largest contour
contours = cv2.findContours(rect, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
contours = contours[0] if len(contours) == 2 else contours[1]
for c in contours:
area_thresh = 0
area = cv2.contourArea(c)
if area > area_thresh:
area = area_thresh
big_contour = c
# get rotated rectangle from contour
rot_rect = cv2.minAreaRect(big_contour)
box = cv2.boxPoints(rot_rect)
box = np.int0(box)
for p in box:
pt = (p[0],p[1])
print(pt)
# draw rotated rectangle on copy of img
rot_bbox = img.copy()
cv2.drawContours(rot_bbox,[box],0,(0,0,255),2)
# write img with red rotated bounding box to disk
cv2.imwrite("rectangle_thresh.png", thresh)
cv2.imwrite("rectangle_outline.png", rect)
cv2.imwrite("rectangle_bounds.png", rot_bbox)
# display it
cv2.imshow("IMAGE", img)
cv2.imshow("THRESHOLD", thresh)
cv2.imshow("RECT", rect)
cv2.imshow("BBOX", rot_bbox)
cv2.waitKey(0)
Thresholded Image:
Rectangle Region Extracted:
Rotated Bounding Box on Image:
Rotated Bounding Box Corners:
(444, 335)
(167, 330)
(170, 120)
(448, 125)
Here's a simple approach:
Obtain binary image. We load the image, grayscale, Gaussian blur, then adaptive threshold.
Morphological operations. We create a rectangular kernel and morph open to remove the small noise
Find distorted rectangle contour and draw onto a mask. Find contours, determine rotated bounding box, and draw onto a blank mask
Find corners. We use the Shi-Tomasi Corner Detector already implemented as cv2.goodFeaturesToTrack which is supposedly shows better results compared to the Harris Corner Detector
Here's a visualization of each step:
Binary image
Morph open
Find rotated rectangle contour and draw/fill onto a blank mask
Draw rotated rectangle and corners to get result
Corner coordinates
(448.0, 337.0)
(164.0, 332.0)
(452.0, 123.0)
(168.0, 118.0)
Code
import cv2
import numpy as np
# Load image, grayscale, Gaussian blur, adaptive threshold
image = cv2.imread("1.png")
mask = np.zeros(image.shape, dtype=np.uint8)
gray = 255 - cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (3,3), 0)
thresh = cv2.adaptiveThreshold(blur, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY_INV, 51, 3)
# Morph open
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3,3))
opening = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel, iterations=1)
# Find distorted rectangle contour and draw onto a mask
cnts = cv2.findContours(opening, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]
rect = cv2.minAreaRect(cnts[0])
box = cv2.boxPoints(rect)
box = np.int0(box)
cv2.drawContours(image,[box],0,(36,255,12),2)
cv2.fillPoly(mask, [box], (255,255,255))
# Find corners on the mask
mask = cv2.cvtColor(mask, cv2.COLOR_BGR2GRAY)
corners = cv2.goodFeaturesToTrack(mask, maxCorners=4, qualityLevel=0.5, minDistance=150)
for corner in corners:
x,y = corner.ravel()
cv2.circle(image,(x,y),8,(255,120,255),-1)
print("({}, {})".format(x,y))
cv2.imshow("thresh", thresh)
cv2.imshow("opening", opening)
cv2.imshow("mask", mask)
cv2.imshow("image", image)
cv2.waitKey(0)
You can try with an adaptive threshold. Then you may either use cornerHarris if you only need corners, or depending on what you need to do next, you could also find useful findContours, which returns a list of bounding boxes
I was able to locate 3 out of the 4 points, the 4th point can be found easily given the other three points since it's rectangle. Here is my solution:
import cv2
import numpy as np
img = cv2.imread('6dUIr.png',1)
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
#smooth the image
kernel = np.ones((5,5),np.float32)/25
gray = cv2.filter2D(gray,-1,kernel)
#histogram equalization
clahe = cv2.createCLAHE(clipLimit=1.45, tileGridSize=(4,4))
cl1 = clahe.apply(gray)
#find edges
edges = cv2.Canny(cl1,4,100)
#find corners
dst = cv2.cornerHarris(edges,2,3,0.04)
#result is dilated for marking the corners, not important
dst = cv2.dilate(dst,None)
# Threshold for an optimal value, it may vary depending on the image.
img[dst>0.25*dst.max()]=[0,0,255]
cv2.imshow('edges', edges)
cv2.imshow('output', img)
# cv2.imshow('Histogram equalized', img_output)
cv2.waitKey(0)
The code has many hard coded thresholds but it's a good start.

Categories