This question already has answers here:
Filling holes inside a binary object
(7 answers)
Closed 11 months ago.
This is my code, I am trying to delete the mask (noise) from the binary image. What I am getting is white lines left around the noise. I am aware that there is a contour around that noise creating the final white line in the results. any help?
Original Image
Mask and results
Code
import numpy as np
import cv2
from skimage import util
img = cv2.imread('11_otsu.png')
imgray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
ret, thresh = cv2.threshold(imgray, 127, 255, 0, cv2.THRESH_BINARY)
contours, hierarchy = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
#cv2.drawContours(img, contours, -1, (0,255,0), 2)
# create an empty mask
mask = np.zeros(img.shape[:2], dtype=np.uint8)
# loop through the contours
for i, cnt in enumerate(contours):
# if the contour has no other contours inside of it
if hierarchy[0][i][2] == -1:
# if the size of the contour is greater than a threshold
if cv2.contourArea(cnt) <70:
cv2.drawContours(mask, [cnt], 0, (255), -1)
# display result
cv2.imshow("Mask", mask)
cv2.imshow("Img", img)
image = cv2.bitwise_not(img, img, mask=mask)
cv2.imshow("Mask", mask)
cv2.imshow("After", image)
cv2.waitKey()
cv2.destroyAllWindows()
Your code is perfectly fine just make these adjustments and it should work:
contours, hierarchy = cv2.findContours(thresh, cv2.RETR_CCOMP, cv2.CHAIN_APPROX_SIMPLE) # Use cv2.CCOMP for two level hierarchy
if hierarchy[0][i][3] != -1: # basically look for holes
# if the size of the contour is less than a threshold (noise)
if cv2.contourArea(cnt) < 70:
# Fill the holes in the original image
cv2.drawContours(img, [cnt], 0, (255), -1)
Instead of trying to find inner contours and filling those in, may I suggest using cv2.floodFill instead? The flood fill operation is commonly used to fill in holes inside closed objects. Specifically, if you set the seed pixel to be the top left corner of the image then flood fill the image, what will get filled is the background while closed objects are left alone. If you invert this image, you will find all of the pixels that are interior to the closed objects that have "holes". If you take this inverted image and use the non-zero locations to directly set the original image, you will thus fill in the holes.
Therefore:
im = cv2.imread('8AdUp.png', 0)
h, w = im.shape[:2]
mask = np.zeros((h+2, w+2), dtype=np.uint8)
holes = cv2.floodFill(im.copy(), mask, (0, 0), 255)[1]
holes = ~holes
im[holes == 255] = 255
cv2.imshow('Holes Filled', im)
cv2.waitKey(0)
cv2.destroyAllWindows()
First we read in the image that you've provided which is thresholded and before the "noise filtering", then get the height and width of it. We also use an input mask to tell us which pixels to operate on the flood fill. Using a mask of all zeroes means you will operate on the entire image. It's also important to note that the mask needs to have a 1 pixel border surrounding it before using it. We flood fill the image using the top left corner as the initial point, invert it, set any "hole" pixels to 255 and show it. Take note that the input image is mutated once the method finishes so you need to pass in a copy to leave the input image untouched. Also, cv2.floodFill (using OpenCV 4) returns a tuple of four elements. I'll let you look at the documentation but you need the second element of this tuple, which is the filled in image.
We thus get:
I think using cv2.GaussianBlur() method might help you. After you convert the image to gray-scale, blur it using this method (as the name suggests, this is a Gaussian filter). Here is the documentation:
https://docs.opencv.org/4.3.0/d4/d86/group__imgproc__filter.html
Related
I wrote a program that takes a picture as an input and detects the written text
I need to edit my program to add two more functionalities
First
since I already was able to get the Contour coordinates(done), I want to know how do I get the color of the background that surrounds the contour line of each character ( at multiple points because I want to calculate the average of RGB color values surrounding that character )
I just don't know if there is already a function to do what I want to do or if there is a specific approach that I should follow
second,
I tried to fill the inside of the Contour but it did not work, can someone help me to know why
I did try both thickness=-1 and thickness=cv2.FILLE in my cv2.drawContours, nothing worked, clearly only the control line is blue but what is inside is not blue
I know that the code is not perfect and there are so many comments but it is because I'm Kim trying new stuff
thank you in advance to everyone who's going to help
import cv2
import pytesseract
import numpy as np
from PIL import ImageGrab
import pyautogui
def Contour(img):
Contours,hierarchy = cv2.findContours(img, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
for cnt in Contours:
area = cv2.contourArea(cnt)
# print(area)
# if area<150: could be useded to be cpecific
# we need a copy of the image so we dont draw on the original or thickness=-1
cv2.drawContours(imgcpy, cnt, -1, (255,0,0), thickness=-1)
#pytesseract.pytesseract.tesseract_cmd= 'C:\\Program Files\\Tesseract-OCR\\tesseract.exe'
cap= cv2
img = cv2.imread('D:\\opencv\\R\\img\\lec\\3.png')
imgcpy = img.copy()
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# for japanese txt = pytesseract.image_to_string(gray, lang='jpn')
txt = pytesseract.image_to_string(gray)
imgblur = cv2.GaussianBlur(img, (7, 7), 0)
imgcanny = cv2.Canny(img, 200, 200)
#imgDialation = cv2.dilate(imgcanny, kernel, iterations=1)
#imgeroded = cv2.erode(imgDialation, kernel, iterations=1)
imBlank = np.zeros_like(img)
# waste of time only the same picture work -_-
# imstack = np.hstack((gray,gray,gray))
Contour(imgcanny)
cv2.imshow('Contour', imgcpy)
#img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
print(txt)
print(Contour(imgcanny))
#cv2.imshow('test', img)
cv2.waitKey()
The "first" issue is solved by
getting a mask, either
using connectedComponents or
by filling a mask using drawContours into np.zeros((height, width), dtype=np.uint8)
dilate the mask, then
subtract the mask from the dilated version (literal subtraction will work, but so do bitwise operations)
The result is a mask that contains pixels just outside of the contour.
Use this mask to get those pixel values:
# have `mask` somehow, which should be dtype uint8
dilated = cv.dilate(mask)
surrounding = dilated - mask
values = img[surrounding != 0] # numpy indexing using a boolean array
meancolor = values.mean(axis=(0,1)) # sum/average over both axes
The "second" issue is that you pass a single contour to drawContours, which expects a list of contours.
Change your code to pass [cnt] instead of just cnt:
cv2.drawContours(imgcpy, [cnt], -1, (255,0,0), thickness=-1)
If you only pass cnt, it will interpret the list of points (a contour is a list of points) as a list of contours, hence it will interpret each point as a contour... and that's just a point, not a contour.
I'm using OpenCv (4.x) on Anime Sketch dataset from Kaggle to get the image's silhouette. What I found to be the hardest part was to detect that empty areas inside that silhouette, areas between arm-body, legs and hair. The tutorials I followed always use "full filled" objects, like a ball, head or cars and I ended up tunning that code to make it work, but it is too specific so that tunning just work ok on one image.
Playing around in online-image-editor.com I've noticed that I can use the tool called Trans-parency to change one color, just like cv2.inRange() does.
Original image
The code:
image = cv2.imread("2.png",cv2.IMREAD_UNCHANGED)
crop_img = image[:, 0:512]
fuzz_factor = 0.97
maxColor = (crop_img[1,1] * 1).astype(int)
minColor = (maxColor * fuzz_factor).astype(int)
mask = cv2.inRange(crop_img, minColor, maxColor)
cv2.imshow("mask", mask)
cv2.waitKey()
and outputs this (not that bad..)
BUT then trying with another image it doesn't work anymore, output:
So, question(s):
There is some "magic rule" where I can extract a specific fuzz_factor for each image?
How could I use the image's right half to get that silhouette/contour?
Thanks guys
I post to close this question.
Thanks to Micka I made some progress, there are two variables that have high impact on output's quality:
fuzz_factor: which sets the color range for cv2.inRange()
max_contours: number of contours to draw (sorted by size)
High numbers are better until there are white zones that are not background, so next thing could be discard that ones.
import numpy as np
import cv2
# constants
fuzz_factor = 1
max_contours = -10
image_path = "9.png"
image = cv2.imread(image_path)
image = image[:, 0:512]
# background color boundaries
color = image[3,3]
upper = (color).astype(int)
lower = (color * (100 - fuzz_factor/2.0)/100).astype(int)
# create mask with specific colors
mask = cv2.inRange(image, lower, upper)
# get all contours
contours, _ = cv2.findContours(mask, mode = cv2.RETR_EXTERNAL, method = cv2.CHAIN_APPROX_NONE)
if(len(contours) > 1):
# get the [max_contours] biggest areas
contours = sorted(contours, key=cv2.contourArea)[max_contours:]
# mask where contours are filled
mask = np.zeros_like(image)
# draw contours and fill
cv2.drawContours(mask, contours, -1, color=[255,255,255], thickness= -1)
cv2.drawContours(image, contours, -1, 255, 2)
cv2.imshow("Result", np.hstack([image, mask]))
cv2.waitKey(0)
I'm doing cell segmentation, so I'm trying to code a function that removes all minor contours around the main one in order to do a mask.
That happens because I load an image with some color markers:
The problem is when I do threshold, it assumes that "box" between the color markers as a part of the main contour.
As you may see in my code, I don't directly pass color image to grays because the red turns black but there are other colors too, at least 8, and always different in each image. I've got thousands of images like this where just one cell is displayed, but in most of it, there are always outsiders contours attached. My goal is to come to a function that gives a binary image of a single cell for each image input like this. So I'm starting with this code:
import cv2 as cv
cell1 = cv.imread(image_cell, 0)
imgray = cv.cvtColor(cell1,cv.COLOR_BGR2HSV)
imgray = cv.cvtColor(imgray,cv.COLOR_BGR2GRAY)
ret,thresh_binary = cv.threshold(imgray,107,255,cv.THRESH_BINARY)
cnts= cv.findContours(image =cv.convertScaleAbs(thresh_binary) , mode =
cv.RETR_TREE,method = cv.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]
for c in cnts:
cv.drawContours(thresh_binary,[c], 0, (255,255,255), -1)
kernel = cv.getStructuringElement(cv.MORPH_RECT, (3,3))
opening = cv.morphologyEx(thresh_binary, cv.MORPH_OPEN, kernel,
iterations=2) # erosion followed by dilation
Summing up, how do I get just the red contour from image 1?
So another approach, without color ranges.
A couple of things are not going right in your code I think. First, you are drawing the contours on thresh_binary, but that already has the outer lines of the other cells as well - the lines you are trying to get rid off. I think that is why you use opening(?) while in this case you shouldn't.
To fix things, first a little information on how findContours works. findContours starts looking for white shapes on a black background and then looks for black shapes inside that white contour and so on. That means that the white outline of the cells in the thresh_binary are detected as a contour. Inside of it are other contours, including the one you want. docs with examples
What you should do is first look only for contours that have no contours inside of them. The findContours also returns a hierarchy of contours. It indicates whether a contour has 'childeren'. If it has none (value: -1) then you look at the size of the contour and disregard the ones that are to small. You could also just look for the largest, as that is probably the one you want. Finally you draw the contour on a black mask.
Result:
Code:
import cv2 as cv
import numpy as np
# load image as grayscale
cell1 = cv.imread("PjMQR.png",0)
# threshold image
ret,thresh_binary = cv.threshold(cell1,107,255,cv.THRESH_BINARY)
# findcontours
contours, hierarchy = cv.findContours(image =thresh_binary , mode = cv.RETR_TREE,method = cv.CHAIN_APPROX_SIMPLE)
# create an empty mask
mask = np.zeros(cell1.shape[:2],dtype=np.uint8)
# loop through the contours
for i,cnt in enumerate(contours):
# if the contour has no other contours inside of it
if hierarchy[0][i][2] == -1 :
# if the size of the contour is greater than a threshold
if cv2.contourArea(cnt) > 10000:
cv.drawContours(mask,[cnt], 0, (255), -1)
# display result
cv2.imshow("Mask", mask)
cv2.imshow("Img", cell1)
cv2.waitKey(0)
cv2.destroyAllWindows()
Note: I used the image you uploaded, your image probably has far fewer pixels, so a smaller contourArea
Note2: enumerate loops through the contours, and returns both a contour and an index for each loop
Actually, in your code the 'box' is a legitimate extra contour. And you draw all contours on the final image, so that includes the 'box'. This could cause issues if any of the other colored cells are fully in the image.
A better approach is to separate out the color you want. The code below creates a binary mask that only displays the pixels that are in the defined range of red colors. You can use this mask with findContours.
Result:
Code:
import cv2
# load image
img = cv2.imread("PjMQR.png")
# Convert HSV
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
# define range of red color in HSV
lower_val = np.array([0,20,0])
upper_val = np.array([15,255,255])
# Threshold the HSV image to get only red colors
mask = cv2.inRange(hsv, lower_val, upper_val)
# display image
cv2.imshow("Mask", mask)
cv2.waitKey(0)
cv2.destroyAllWindows()
This code can help you understand how the different values in this process (HSV with inRange) works. inRange docs
I'm in the process of scanning old photographs, and I would like to automate the process of extracting the photograph from the (noisy) solid white background of the scanner so that I have a transparent photograph. This part of the program now works, but I have one more small problem with this.
The photograph can now be accurately detected (and extracted), but it leaves a small and sharp black border from the background around the entire photograph. I've tried to apply a gaussian blur to the transparency mask, but this wouldn't smooth the blackness away (and it made the border of the photograph look 'smudged').
This is the code that I have to extract the photo and generate the transparency mask:
# Load the scan, and convert it to RGBA.
original = cv2.imread('input.jpg')
original = cv2.cvtColor(original, cv2.COLOR_BGR2BGRA)
# Make the scan grayscale, and apply a blur.
image = cv2.cvtColor(original, cv2.COLOR_BGR2GRAY)
image = cv2.GaussianBlur(image, (25, 25), 0)
# Binarize the scan.
retval, threshold = cv2.threshold(image, 50, 255, cv2.THRESH_BINARY)
# Find the contour of the object.
contours, hierarchy = cv2.findContours(threshold, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
largestContourArea = -1
largestContour = -1
for contour in contours:
area = cv2.contourArea(contour)
if area > largestContourArea:
largestContourArea = area
largestContour = contour
# Generate the transparency mask.
mask = numpy.zeros(original.shape, numpy.uint8)
cv2.drawContours(mask, [ largestContour ], -1, (255, 255, 255, 255), -1)
# Apply the transparency mask.
original = cv2.multiply(mask.astype(float) / 255.0, original.astype(float))
cv2.imwrite('output.png', original)
I have a sample scan and the result of the code above using the sample scan. As you can see, there is a slight black border all around the photograph, which I would like to remove.
By using the erode method, you could shrink the contour (mask), effectively removing the black edge.
Since this method supports in-place operation, the code would look something like this: cv2.erode(mask, mask, kernel), where kernel is a kernel acquired using cv2.getStructuringElement. By playing around with the kernel and the iteration count, you can control the amount of shrinking.
This page explains everything in great detail and also provides nice examples: https://docs.opencv.org/2.4/doc/tutorials/imgproc/erosion_dilatation/erosion_dilatation.html
So I decided to get started learning Open CV and Python together!
My first project is to detect moving objects on a relatively still background and then detect their average color to sort them. There are at least 10 objects to detect and I am processing a colored video.
So far I managed to remove the background, identify the contours (optionally get the center of each contour) but now I am struggling getting the average or mean color inside of each contour. There are some topics about this kind of question but most of them are written in C. Apparently I could use cv.mean() but I can't get a working mask to feed in this function. I guess it's not so difficult but I am stuck there... Cheers!
import numpy as np
import cv2
video_path = 'test.h264'
cap = cv2.VideoCapture(video_path)
fgbg = cv2.createBackgroundSubtractorMOG2()
while (cap.isOpened):
ret, frame = cap.read()
if ret==True:
fgmask = fgbg.apply(frame)
(contours, hierarchy) = cv2.findContours(fgmask, cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)
for c in contours:
if cv2.contourArea(c) > 2000:
cv2.drawContours(frame, c, -1, (255,0,0), 3)
cv2.imshow('foreground and background',fgmask)
cv2.imshow('rgb',frame)
key = cv2.waitKey(1) & 0xFF
if key == ord("q"):
break
cap.release()
cv2.destroyAllWindows()
You can create a mask by first creating a new image with the same dimensions as your input image and pixel values set to zero.
You then draw the contour(s) onto this image with pixel value 255. The resulting image can be used as a mask.
mask = np.zeros(frame.shape, np.uint8)
cv2.drawContours(mask, c, -1, 255, -1)
mask can then be used as a parameter to cv.mean like
mean = cv.mean(frame, mask=mask)
Just one word of caution, the mean of RGB colors does not always make sense. Maybe try converting to HSV color space and solely use the H channel for detecting the color of your objects.
Solution on an image
1) find contour (in this case rectangle, contour that is not rectangle is much harder to make)
2) find coordiantes of contour
3) cut the image from contour
4) sum individual channels and divide them by number of pixels in it ( or with mean function)
import numpy as np
import cv2
img = cv2.imread('my_image.jpg',1)
cp = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
ret,thresh = cv2.threshold(cp,150,255,0)
cv2.imshow('img',thresh)
cv2.waitKey(0)
im2,contours,hierarchy = cv2.findContours(thresh.astype(np.uint8), 1, 2)
cnts = contours
for cnt in cnts:
if cv2.contourArea(cnt) >800: # filter small contours
x,y,w,h = cv2.boundingRect(cnt) # offsets - with this you get 'mask'
cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),2)
cv2.imshow('cutted contour',img[y:y+h,x:x+w])
print('Average color (BGR): ',np.array(cv2.mean(img[y:y+h,x:x+w])).astype(np.uint8))
cv2.waitKey(0)
cv2.imshow('img',img)
cv2.waitKey(0)
cv2.destroyAllWindows()
To remove noise, you can just take center of the contour, and take smaller rectangle to examin.
For non rectangular contour, look at cv2.fillPoly function -> Cropping non rectangular contours. But its a bit slow algorithm (but nothing limiting)
If you are interested in non rectangular contour, you will have to be careful about doing mean, because you will need mask and the mask/background is always rectangular so you will be doing mean on something you dont want