I have the following image
lower = np.array([175, 125, 45], dtype="uint8")
upper = np.array([255, 255, 255], dtype="uint8")
mask = cv2.inRange(image, lower, upper)
img = cv2.bitwise_and(image, image, mask=mask)
plt.figure()
plt.imshow(img)
plt.axis('off')
plt.show()
now if I try to transform into grayscale like this:
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
I get that:
And I would like to extract the number on it.
The suggestion:
gray = 255 - gray
emp = np.full_like(gray, 255)
emp -= gray
emp[emp==0] = 255
emp[emp<100] = 0
gauss = cv2.GaussianBlur(emp, (3,3), 1)
gauss[gauss<220] = 0
plt.imshow(gauss)
gives the image:
Then using pytesseract on any of the images:
data = pytesseract.image_to_string(img, config='outputbase digits')
gives:
'\x0c'
Another suggested solution is:
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
thr = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY_INV)[1]
txt = pytesseract.image_to_string(thr)
plt.imshow(thr)
And this gives
'\x0c'
Not very satisfying... Anyone has a better solution please?
Thanks!
I have a two step solution
Apply thresholding
Set psm mode to 7.
When you apply thresholding to the image:
Thresholding is a simplest method of displaying the features of the image.
Now from the output image, when we read:
txt = image_to_string(thr, config="--psm 7")
print(txt)
Result will be:
| 1,625 |
Now why do we set page-segmentation-mode (psm) mode to the 7?
Well, treating image as a single text line will give the accurate result.
But we have to modify the result. Since the current result is | 1,625 |
We should remove the |
print("".join([t for t in txt if t != '|']))
Result:
1,625
Code:
import cv2
from pytesseract import image_to_string
img = cv2.imread("LZ3vi.png")
gry = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
thr = cv2.threshold(gry, 0, 255,
cv2.THRESH_BINARY_INV)[1]
txt = image_to_string(thr, config="--psm 7")
print("".join([t for t in txt if t != '|']).strip())
Update
how do you get this clean black and white image from my original image?
Using 3-steps
Reading the image using opencv's imread function
img = cv2.imread("LZ3vi.png")
Now we read the image in BGR fashion. (Not RGB)
Convert the image to the graysclae
gry = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
Result will be:
Apply threshold
thr = cv2.threshold(gry, 0, 255, cv2.THRESH_BINARY_INV)[1]
Result will be:
Now if you are wondering about thresholding. Read the simple-threhsolding
All my filters, grayscale... get weird colored images
The reason is, when you are displaying the image using pyplot, you need to set color-map (cmap) to gray
plt.imshow(img, cmap='gray')
You can read the other types here
Two issues blocked the pytessract from detecting your number:
The white rectangle around the number(Inverting and filling is the solution).
The Noise in the numbers shape(Gaussian Smoothing dealt with that)
The solution that AlexAlex has proposed will work perfectly if it was followed by a Gaussian filter:
output: 1,625
import numpy as np
import pytesseract
import cv2
BGR = cv2.imread('11.png')
RGB = cv2.cvtColor(BGR, cv2.COLOR_BGR2RGB)
lower = np.array([175, 125, 45], dtype="uint8")
upper = np.array([255, 255, 255], dtype="uint8")
mask = cv2.inRange(RGB, lower, upper)
img = cv2.bitwise_and(RGB, RGB, mask=mask)
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
gray = 255 - gray
emp = np.full_like(gray, 255)
emp -= gray
emp[emp==0] = 255
emp[emp<100] = 0
gauss = cv2.GaussianBlur(emp, (3,3), 1)
gauss[gauss<220] = 0
text = pytesseract.image_to_string(gauss, config='outputbase digits')
print(text)
Related
I need to find several objects (let's say up to 10 pieces) in the picture. Objects must be of a certain color.
Using this code and searching for contours, it turns out to find only the midpoint between several objects, but you need to find the coordinates of each object separately.
Maybe somehow you can make a limit on the size of one object? (i.e. the object must not be less than 100x100 pixels, and not more than 300x300 pixels)
import cv2
import matplotlib.pyplot as plt
import pyautogui as pg
output = '1.jpg'
flags = [i for i in dir(cv2) if i.startswith('COLOR_')]
nemo = cv2.imread(output, cv2.IMREAD_COLOR) #output, cv2.IMREAD_COLOR)
nemo = cv2.cvtColor(nemo, cv2.COLOR_BGR2RGB)
hsv_nemo = cv2.cvtColor(nemo, cv2.COLOR_RGB2HSV)
light_orange = (92,128,224)
dark_orange = (109,255,255)
#
mask = cv2.inRange(hsv_nemo, light_orange, dark_orange)
result = cv2.bitwise_and(nemo, nemo, mask=mask)
#
moments = cv2.moments(mask, 1)
x_moment = moments['m01']
y_moment = moments['m10']
area = moments['m00']
y = int(x_moment / area)
x = int(y_moment / area)
pg.moveTo(x,y)
plt.imshow(result)
plt.imsave('result.jpg',result)
plt.show()
#print(x,y)
gray = cv2.cvtColor(result, cv2.COLOR_BGR2GRAY)
ret, binary = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY)
contours, hierarchy = cv2.findContours(binary, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cv2.drawContours(result, contours, -1, (0, 0, 255), 3)
cv2.imshow('img', result)
cv2.waitKey(0)
#print (len(contours[0]))
#print (len(contours[1]))
enter image description here
I have to crop a lot of images manually. Not the funniest thing to do. So I thought I'd try to do it using Python.
I can detect the subject, create a mask, but I have no idea how to get the points from the very bottom part and crop based on them.
Any help is appreciated
import cv2
img = cv2.imread('image5.jpg')
h, w = img.shape[:2]
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
thr = cv2.threshold(gray, 192, 255, cv2.THRESH_BINARY_INV)[1]
cv2.imwrite('result5.png', thr)
you can try to find all external contours using cv2.RETR_EXTERNAL and pick the bottom most point, like this:
import cv2
import numpy as np
import imutils
im = cv2.imread('images/tennis.jpg')
# Percent of original size
scale_percent = 20
width = int(im.shape[1] * scale_percent / 100)
height = int(im.shape[0] * scale_percent / 100)
dim = (width, height)
# Resize image
im = cv2.resize(im, dim, interpolation = cv2.INTER_AREA)
# Convert to grayscale
gray = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
# Canny
canny_output = cv2.Canny(im, 120, 240)
# Find external contours
contours, hierarchy = cv2.findContours(canny_output, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
#cv2.drawContours(im, [contours[0]], 0, (0,255,0), 3) # Uncomment this line to see what contour opencv is finding
# Pick the bottom most point and add an offset (whatever value you want, this is just for aesthetics)
c = contours[0]
bottommost = tuple(c[c[:, :, 1].argmax()][0])[1] + 5
# Crop image
im = im[:bottommost, :]
# Show image
cv2.imshow('image', im)
cv2.waitKey()
Very good thinking I'd say! now the implementation:
xx,yy = thrs.nonzero()
max_crop_h = xx.max()
crop = img[:max_crop_h,:]
numpy has your back!
I need to read the highest temperature on thermographic images, as shown below:
IR_1544_INFRA.jpg
IR_1546_INFRA.jpg
IR_1560_INFRA.jpg
IR_1564_INFRA.jpg
I used the following code, this was the best result.
I also tried several other ways, such as: blur, gray scale, binarization, and others but they all failed.
import cv2
import pytesseract
pytesseract.pytesseract.tesseract_cmd = r"C:\Users\User\AppData\Local\Tesseract-OCR\tesseract.exe"
# Load image, grayscale, Otsu's threshold
entrada = cv2.imread('IR_1546_INFRA.jpg')
image = entrada[40:65, 277:319]
#image = cv2.imread('IR_1546_INFRA.jpg')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
thresh = 255 - cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
# Blur and perform text extraction
thresh = cv2.GaussianBlur(thresh, (3,3), 0)
data = pytesseract.image_to_string(thresh, lang='eng', config='--psm 6')
print(data)
cv2.imshow('thresh', thresh)
cv2.waitKey()
In the first image, I found
this
In the second image, I found this.
The imagem layout is always the same, that is, the temperature is always in the same place, so I cropped the image to isolate only the number. I would like (97.7 here, and 85.2 here).
My code needs to find from these images to always detect this temperature and generate a list indicating from highest to lowest.
What do you indicate for me to improve the assertiveness of pytesseract in the case of these images?
Note 1: When I annalyze the entire image (without cropping), it returns data that is not even present.
Note 2: In some images even with the binary number, pytesseract (image_to_string) does not return any data.
Thank you all and sorry for the typos, writing in english is still a challenge for me.
Because you have same images, you can crop the area you want and then do processing there. The processing is also simple. Change to gray, get threshold, invert, resize, and then do the OCR. You can see it in my code below. It works on all your attached images.
import cv2
import pytesseract
import os
image_path = "temperature"
for nama_file in sorted(os.listdir(image_path)):
print(nama_file)
img = cv2.imread(os.path.join(image_path, nama_file))
crop = img[43:62, 278:319]
gray = cv2.cvtColor(crop, cv2.COLOR_BGR2GRAY)
thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY)[1]
thresh = cv2.bitwise_not(thresh)
double = cv2.resize(thresh, None, fx=2, fy=2)
custom_config = r'-l eng --oem 3 --psm 7 -c tessedit_char_whitelist="1234567890." '
text = pytesseract.image_to_string(double, config=custom_config)
print("detected: " + text)
cv2.imshow("img", img)
cv2.imshow("double", double)
cv2.waitKey(0)
cv2.destroyAllWindows()
I am new to image processing and python. You might've seen my amateur codes on this site in the last couple of days.
I am trying to count the number of trees using aerial images. This is my code:
from PIL import Image
import cv2
import numpy as np
from skimage import io, filters, measure
from scipy import ndimage
img = Image.open("D:\\Texture analysis\\K-2.jpg")
row, col = img.size
hsvimg = img.convert('HSV')
hsvimg.mode = 'RGB'
hsvimg.save('newImage2.jpg')
npHSI = np.asarray(hsvimg) #Convert HSI Image to np image
blur = cv2.GaussianBlur(npHSI, (45, 45), 5)
assert isinstance(blur, np.ndarray) ##############################
assert len(blur.shape) == 3 #Convert np Image to HSI Image
assert blur.shape[2] == 3 ##############################
hsiBlur = Image.fromarray(blur, 'RGB')
hsiBlur.save('hsiBlur.jpg') #Save the blurred image
## Read
img = cv2.imread("D:\\Texture analysis\\hsiBlur.jpg")
## convert to hsv
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
#Threshold the image and segment the trees
mask = cv2.inRange(hsv, (36, 25, 25), (70, 255,255))
imask = mask>0
green = np.zeros_like(img, np.uint8)
green[imask] = img[imask]
## save
cv2.imwrite("green.png", green)
#Count the number of trees
im = io.imread('green.png', as_grey=True)
val = filters.threshold_otsu(im)
drops = ndimage.binary_fill_holes(im < val)
labels = measure.label(drops)
print(labels.max())
Original image:
HSI image with gaussian filter:
Segmented image:
The last part of the code returns 7, which is a wrong output. The value should be above 50. How can I properly count the number of green segments in the final segmented image?
EDIT
I converted green.png to binary and applied erosion with a 3x3 filter and iterated it 7 times to remove the noise.
This is what I did at the end. I followed this stackoverflow link
##save
cv2.imwrite("green.png", green)
#Convert to grayscale
gray = np.dot(green[...,:3], [0.299, 0.587, 0.114])
cv2.imwrite("grayScale.jpg", gray)
#Binarize the grayscale image
ret,bin_img = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)
cv2.imwrite("bin_img.jpg", bin_img)
#Erosion to remove the noise
kernel = np.ones((3, 3),np.uint8)
erosion = cv2.erode(gray, kernel, iterations = 7)
cv2.imwrite("erosion.jpg", erosion)
#Count the number of trees
finalImage = cv2.imread('erosion.jpg')
finalImage = cv2.cvtColor(finalImage, cv2.COLOR_BGR2GRAY)
ret, thresh = cv2.threshold(finalImage, 127, 255, 1)
im2, contours, hierarchy = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
for cnt in contours:
cv2.drawContours(finalImage,[cnt],0,(0,0,255),1)
Saurav mentioned in his answer ... size of "contours" will give you the count. This print(contour.size())gives an error and print(contour) just prints a long 2D array. How can i get the size of contour?
PS. I didn't upload the grayscale, binary and eroded image because i felt that the images were already taking too much space, I can still upload them if anyone wants to.
I've found 52 trees with that script:
from PIL import Image, ImageDraw, ImageFont
image = Image.open('04uX3.jpg')
pixels = image.load()
size = image.size
draw = ImageDraw.Draw(image)
font = ImageFont.truetype('arial', 60)
i = 1
for x in range(0, size[0], 100):
for y in range(0, size[1], 100):
if pixels[x, y][1] > 200:
draw.text((x, y), str(i), (255, 0, 0), font=font)
i += 1
image.save('result.png')
You can see that some trees weren't detected and some non-trees were detected. So this is very rough calculation:
Like I converted my original input to HSV color space image & applied the INRANGE function and found the green & blue lines & now i want to get rid of them and I want the image to look like in output....how shall i now get rid of the lines & replace them by the background color??
Code Snippet:
import cv2 as cv
import numpy as np
img= cv.imread('C:\input.png',1)
hsv=cv.cvtColor(img,cv.COLOR_BGR2HSV)
lower_green = np.array([30,70,20])
upper_green = np.array([70,255,255])
lower_blue = np.array([95, 110, 20])
upper_blue = np.array([135, 255, 255])
mask = cv.inRange(hsv, lower_green , upper_blue)
res = cv.bitwise_and(img,img, mask= mask)
cv.imwrite("out2.jpg", res)
Here is a quick and dirty solution.
Create a mask from manually threshold image containing the lines (mask 1)
Also create a binary inverted image of this mask (mask 2)
Mask the image of the shirt with mask 1
Inpaint the image above using mask 2
The solution definitely can be improved by performing morphological operations on the mask to remove the lines. Share your thoughts as well
By doing something like #jeru-luke said, the result will be like this:
import cv2 as cv
import numpy as np
img = cv.imread('z12.png', 1)
hsv = cv.cvtColor(img, cv.COLOR_BGR2HSV)
gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
lower_green = np.array([30, 70, 20])
upper_green = np.array([70, 255, 255])
lower_blue = np.array([95, 110, 20])
upper_blue = np.array([135, 255, 255])
mask = cv.inRange(hsv, lower_green, upper_blue)
mask = cv.bitwise_not(mask)
bk = np.full(img.shape, 255, dtype=np.uint8) # white bk
fg_masked = cv.bitwise_and(img, img, mask=mask)
# get masked background, mask must be inverted
mask = cv.bitwise_not(mask)
bk_masked = cv.bitwise_and(bk, bk, mask=mask)
# combine masked foreground and masked background
final = cv.bitwise_or(fg_masked, bk_masked)
cv.imwrite('out_put.png', final)
cv.imshow('final', final), cv.waitKey(0)
import cv2 as cv
import numpy as np
img= cv.imread(r'input.png',1)
hsv=cv.cvtColor(img,cv.COLOR_BGR2HSV)
h,s,v = cv.split(hsv)
th, threshed = cv.threshold(s, 100, 255, cv.THRESH_OTSU|cv.THRESH_BINARY) #black background
mask_w = cv.bitwise_not(threshed) #white background
fg_masked = cv.bitwise_and(v, v, mask=mask_w) #masking the image of shirt with mask_w
dst = cv.inpaint(fg_masked,threshed,3, cv.INPAINT_NS) #inpainting
#Dilation & Erosion.
kernel = np.ones((4, 4),np.uint8)
dilation = cv.dilate(dst,kernel,iterations = 2)
erosion = cv.erode(dilation, kernel, iterations=1)
dilation2= cv.dilate(erosion,kernel,iterations = 1)
dilation3= cv.dilate(dilation2,kernel,iterations = 1)
erosion_final = cv.erode(dilation3, kernel, iterations=3)
cv.imwrite("output_2 [improved].png", erosion_final)