I'm writing an Android app in OpenCV to detect blobs. One task is to threshold the image to differentiate the foreground objects from the background (see image).
It works fine as long as the image is known and I can manually pass a threshold value to threshold()--in this particular image say, 200. But assuming that the image is not known with the only knowledge that there would be a dark solid background and lighter foreground objects how can I dynamically figure out the threshold value?
I've come across the histogram where I can compute the intensity distribution of the grayscale image. But I couldn't find a method to analyze the histogram and choose the value where the objects of interest (lighter) lies. That is; I want to differ the obviously dark background spikes from the lighter foreground spikes--in this case above 200, but in another case could be say, 100 if the objects are grayish.
If all your images are like this, or can be brought to this style, i think cv2.THRESHOLD_OTSU, ie otsu's tresholding algorithm is a good shot.
Below is a sample using Python in command terminal :
>>> import cv2
>>> import numpy as np
>>> img2 = cv2.imread('D:\Abid_Rahman_K\work_space\sofeggs.jpg',0)
>>> ret,thresh = cv2.threshold(img2,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
>>> ret
122.0
ret is the threshold value which is automatically calculated. We just pass '0' as threshold value for this.
I got 124 in GIMP ( which is comparable to result we got). And it also removes the noise. See result below:
If you say that the background is dark (black) and the foreground is lighter, then I recommend to use the YUV color space (or any other YXX like YCrCb, etc.), because the first component of such color spaces is luminance (or lightning).
So after the Y channel is extracted (via the extractChennel function) we need to analyse the histogram of this channel (image):
See the first (left) hump? It represents dark areas (the background in your situation) on your image. So our aim now is to find a segment (on abscissa, it's red part in the image) that contains this hump. Obviously the left point of this segment is zero. The right point is the first point where:
the (local) maximum of histogram is from the left of the point
the value of histogram is less than some small epsilon (you can set it to 10)
I drew a green vertical line to show the location of the right point of the segment in this histogram.
And that's it! This right point of the segment is the needed threshold. Here's the result (epsilon is 10 and the calculated threshold is 50):
I think that it's not a problem for you to delete the noise in the image above.
The following is a C++ implementation of Abid's answer that works with OpenCV 3.x:
// Convert the source image to a 1 channel grayscale:
Mat gray;
cvtColor(src, gray, CV_BGR2GRAY);
// Apply the threshold function with the CV_THRESH_OTSU setting as well
// You can skip having it return the value, but I include it for showing the
// results from OTSU
double thresholdValue = threshold(gray, gray, 0, 255, CV_THRESH_BINARY+CV_THRESH_OTSU);
// Present the threshold value
printf("Threshold value: %f\n", thresholdValue);
Running this against the original image, I get the following:
OpenCV calculated a threshold value of 122 for it, close to the value Abid found in his answer.
Just to verify, I altered the original image as seen here:
And produced the following, with a new threshold value of 178:
Related
I came across this Kaggle kernel that has the following function.
def subtract_gaussian_blur(img):
gb_img = cv2.GaussianBlur(img, (0, 0), 5)
return cv2.addWeighted(img, 4, gb_img, -4, 128)
That converts this RGB image.
Into the following image.
I can see the effect is that it somewhat sharpens the image and turns it into a more grayscale image (not actually grayscale since the image is still RGB) but I'm not actually sure I fully understand what is happening in the function even after reading the OpenCV docs on GaussianBlur and addWeighted.
Also, does this particular image transformation have a specific name that I can do further reading into?
The main step I can see is cv2.addWeighted(img, 4, gb_img, -4, 128). The underlying equation for addWeighted is dst(I)=saturate(src1(I)∗alpha+src2(I)∗beta+gamma). In the example here, alpha is 4, beta -4, and gamma 128.
My understanding of how that works is it first performs a gaussian blur to make a denoised version of the image. However as well as removing noise, Gaussian Blurring can also "smear" edges, which is important later. It then subtracts the denoised version from the original, and adds 128 to each pixel colour channel.
In regions where the original pixel is identical to the filtered pixel, this will result in a uniform grey region. In areas where the original and filtered pixels differ a lot, you will end up either with a lighter or darker region depending on whether the intensity of the original or filtered pixel is higher. The differences will be most pronounced around edges in the original image, because those will be strongly "smeared" by the gaussian blur.
The result isn't fully greyscale as addWeighted() is applied to each colour channel of the pixels separately. Areas where the RGB values of the pre and post blur images differ in an unbalanced way (ie the difference between the two red channels is much bigger than between the blue or green channels) there will be a degree of colour rather than just grey.
I apologize in advance if a question like this was already answered. All of my searches for adding filters resulted in how to add dog faces. I wasn't sure what the proper terminology is.
What techniques do phone apps (such as Snapchat's text overlay or a QR code program for android) use to "darken" a section of the image? I am looking to replace this functionality in OpenCV. Is it possible to do this with other colors? (Such as adding a blue accent)
Example: Snapchat text overlay
https://i.imgur.com/9NHfiBY.jpg
Another Example: Google Allo QR code search
https://i.imgur.com/JnMzvWT.jpg
Any questions or comments would be appreciated
In General:
Change of brightness can be achieved via Addition/Subtraction.
If you want to brighten your Image, you can add a specific amount (e.g. 20) to each channel of the image. The other way around for darkening (Subtraction).
If you would subtract 50 from each channel of the image, you would get:
To darken pixel dependent you could also use Division. This is how a division with 1.5 would change the image:
Another way would be to use the Exponential Operator. This operator takes the value of each channel and will then calculate pixel^value. The resulting value will be then scaled back to the 0-255 range (for 8 bit RGB) via looking the highest value and then calculating the scaling factor via 255/resulting value.
If use it with values > one, it will darker the image. This is because
Here a chart how the exponential operator will change the value of each pixel. As you can see, values for the operator above 1 will darken the image (meaning the channels will be shifted towards lower values), whilst values below 0 will shift all pixels towards white and thus increase brightness.
Here is an example image for application of the operator using the value 0.5, meaning you take each pixel^0.5 and scale it back to the range of 0-255:
For a value of 2 you get:
Sadly i can not help you further, because i am not familiar with OpenCV, but it should be easy enough to implement yourself.
For your question about tinting: Yes, that is also possible. Instead of shifting towards white, you would have to shift the values of each pixel towards the respective color. I recommend to inform you about blending.
Original image taken from here
Update: I was able to darken an image by blending an image matrix with a black matrix. After that, it was just a matter of darkening certain parts of the image to replicate an overlay.
The lower the alpha value is, the darker the image.
Result
void ApplyFilter(cv::Mat &inFrame, cv::Mat &outFrame, double alpha)
{
cv::Mat black = cv::Mat(inFrame.rows, inFrame.cols, inFrame.type(), 0.0);
double beta = (1.0 - alpha);
cv::addWeighted(inFrame, alpha, black, beta, 0.0, outFrame);
}
https://docs.opencv.org/2.4/doc/tutorials/core/adding_images/adding_images.html
Thank you for the help everyone!
I wanna calculate the perceived brightness of an image and classify the image into dark, neutral and bright. And I find one problem here!
And I quote Lakshmi Narayanan's comment below. I'm confused with this method. What does "the average of the hist values from 0th channel" mean here? the 0th channel refer to gray image or value channel in hsv image? Moreover, what's the theory of that method?
Well, for such a case, I think the hsv would be better. Or try this method #2vision2. Compute the laplacian of the gray scale of the image. obtain the max value using minMacLoc. call it maxval. Estimate your sharpness/brightness index as - (maxval * average V channel values) / (average of the hist values from 0th channel), as said above. This would give you certain values. low bright images are usually below 30. 30 - 50 can b taken as ok images. and above 50 as bright images.
If you have an RGB color image you can get the brightness by converting it to another color space that separates color from intensity information like HSV or LAB.
Gray images already show local "brightness" so no conversion is necessary.
If an image is perceived as bright depends on many things. Mainly your display device, reference images, contrast, human...
Using a few intensity statistics values should give you an ok classification for one particular display device.
Given an image (Like the one given below) I need to convert it into a binary image (black and white pixels only). This sounds easy enough, and I have tried with two thresholding functions. The problem is I cant get the perfect edges using either of these functions. Any help would be greatly appreciated.
The filters I have tried are, the Euclidean distance in the RGB and HSV spaces.
Sample image:
Here it is after running an RGB threshold filter. (40% it more artefects after this)
Here it is after running an HSV threshold filter. (at 30% the paths become barely visible but clearly unusable because of the noise)
The code I am using is pretty straightforward. Change the input image to appropriate color spaces and check the Euclidean distance with the the black color.
sqrt(R*R + G*G + B*B)
since I am comparing with black (0, 0, 0)
Your problem appears to be the variation in lighting over the scanned image which suggests that a locally adaptive thresholding method would give you better results.
The Sauvola method calculates the value of a binarized pixel based on the mean and standard deviation of pixels in a window of the original image. This means that if an area of the image is generally darker (or lighter) the threshold will be adjusted for that area and (likely) give you fewer dark splotches or washed-out lines in the binarized image.
http://www.mediateam.oulu.fi/publications/pdf/24.p
I also found a method by Shafait et al. that implements the Sauvola method with greater time efficiency. The drawback is that you have to compute two integral images of the original, one at 8 bits per pixel and the other potentially at 64 bits per pixel, which might present a problem with memory constraints.
http://www.dfki.uni-kl.de/~shafait/papers/Shafait-efficient-binarization-SPIE08.pdf
I haven't tried either of these methods, but they do look promising. I found Java implementations of both with a cursory Google search.
Running an adaptive threshold over the V channel in the HSV color space should produce brilliant results. Best results would come with higher than 11x11 size window, don't forget to choose a negative value for the threshold.
Adaptive thresholding basically is:
if (Pixel value + constant > Average pixel value in the window around the pixel )
Pixel_Binary = 1;
else
Pixel_Binary = 0;
Due to the noise and the illumination variation you may need an adaptive local thresholding, thanks to Beaker for his answer too.
Therefore, I tried the following steps:
Convert it to grayscale.
Do the mean or the median local thresholding, I used 10 for the window size and 10 for the intercept constant and got this image (smaller values might also work):
Please refer to : http://homepages.inf.ed.ac.uk/rbf/HIPR2/adpthrsh.htm if you need more
information on this techniques.
To make sure the thresholding was working fine, I skeletonized it to see if there is a line break. This skeleton may be the one needed for further processing.
To get ride of the remaining noise you can just find the longest connected component in the skeletonized image.
Thank you.
You probably want to do this as a three-step operation.
use leveling, not just thresholding: Take the input and scale the intensities (gamma correct) with parameters that simply dull the mid tones, without removing the darks or the lights (your rgb threshold is too strong, for instance. you lost some of your lines).
edge-detect the resulting image using a small kernel convolution (5x5 for binary images should be more than enough). Use a simple [1 2 3 2 1 ; 2 3 4 3 2 ; 3 4 5 4 3 ; 2 3 4 3 2 ; 1 2 3 2 1] kernel (normalised)
threshold the resulting image. You should now have a much better binary image.
You could try a black top-hat transform. This involves substracting the Image from the closing of the Image. I used a structural element window size of 11 and a constant threshold of 0.1 (25.5 on for a 255 scale)
You should get something like:
Which you can then easily threshold:
Best of luck.
I have been trying to obtain the image brightness in Opencv, and so far I have used calcHist and considered the average of the histogram values. However, I feel this is not accurate, as it does not actually determine the brightness of an image. I performed calcHist over a gray scale version of the image, and tried to differentiate between the avergae values obtained from bright images over that of moderate ones. I have not been successful so far. Could you please help me with a method or algorithm, that can be realised through OpenCv, to estimate brightness of an image? Thanks in advance.
I suppose, that HSV color model will be usefull in your problem, where channel V is Value:
"Value is the brightness of the color and varies with color saturation. It ranges from 0 to 100%. When the value is ’0′ the color space will be totally black. With the increase in the value, the color space brightness up and shows various colors."
So use OpenCV method cvCvtColor(const CvArr* src, CvArr* dst, int code), that converts an image from one color space to another. In your case code = CV_BGR2HSV.Than calculate histogram of third channel V.
I was about to ask the same, but then found out, that similar question gave no satisfactory answers. All answers I've found on SO deal with human observation of a single pixel RGB vs HSV.
From my observations, the subjective brightness of an image also depends strongly on the pattern. A star in a dark sky may look more bright than a cloudy sky by day, while the average pixel value of the first image will be much smaller.
The images I use are grey-scale cell-images produced by a microscope. The forms vary considerably. Sometimes they are small bright dots on very black background, sometimes less bright bigger areas on not so dark background.
My approach is:
Find histogram maximum (HMax) using threshold for removing hot pixels.
Calculate mean values of all pixel between HMax * 2/3 and HMax
The ratio 2/3 could be also increased to 3/4 (which reduces the range of pixels considered as bright).
The approach works quite well, as different cell-patterns with same titration produce similar brightness.
P.S.: What I actually wanted to ask is, whether there is a similar function for such a calculation in OpenCV or SimpleCV. Many thanks for any comments!
I prefer Valentin's answer, but for 'yet another' way of determining average-per-pixel brightness, you can use numpy and a geometric mean instead of arithmetic. To me it has better results.
from numpy.linalg import norm
def brightness(img):
if len(img.shape) == 3:
# Colored RGB or BGR (*Do Not* use HSV images with this function)
# create brightness with euclidean norm
return np.average(norm(img, axis=2)) / np.sqrt(3)
else:
# Grayscale
return np.average(img)
A bit of OpenCV C++ source code for a trivial check to differentiate between light and dark images. This is inspired by the answer above provided years ago by #ann-orlova:
const int darkness_threshold = 128; // you need to determine what threshold to use
cv::Mat mat = get_image_from_device();
cv::Mat hsv;
cv::cvtColor(mat, hsv, CV_BGR2HSV);
const auto result = cv::mean(hsv);
// cv::mean() will return 3 numbers, one for each channel:
// 0=hue
// 1=saturation
// 2=value (brightness)
if (result[2] < darkness_threshold)
{
process_dark_image(mat);
}
else
{
process_light_image(mat);
}