GaussianBlur that fit the whole image - python

I have to filter high frequencies from the input image and I know that Gaussian filter is good in this task. First I transform my image:
import cv2
import numpy as np
from numpy.fft import fft2, ifft2, fftshift, ifftshift
...
image = fftshift(fft2(image))
image_real = image.real
image_imag = image.imag
get the shape of image, blur it and combine
x, y = image.shape
image_real_filter = cv2.GaussianBlur(image_real, (x, y), 0)
image_imag_filter = cv2.GaussianBlur(image_imag, (x, y), 0)
image = image_real_filter + 1j * image_imag_filter
And transform it back.
image = ifft2(ifftshift(image))
It should filter high frequencies but it doesn't I get the black image.

Related

Image processing for blur detection

I am trying to detect blurred images. Thanks to this post, I managed to create a script using Fast Fourier Transform and so far it worked quite well. But for some photos, I am not able to get correct results.
When the background is almost as the same color than the objects from the front, I think my script is not able to give good result.
Do you have any leads to correct this ?
import cv2
import imutils
from PIL import Image as pilImg
from IPython.display import display
import numpy as np
from matplotlib import pyplot as plt
def detect_blur_fft(image, size=60, thresh=17, vis=False):
"""
Detects blur by comparing the image to a blurred version of the image
:param image: The image to detect blur in
:param size: the dimension of the smaller square extracted from the image, defaults to 60 (optional)
:param thresh: the lower this value, the more blur is acceptable, defaults to 17 (optional)
:param vis: Whether or not to return a visualization of the detected blur points, defaults to False
(optional)
"""
# grab the dimensions of the image and use the dimensions to
# derive the center (x, y)-coordinates
(h, w) = image.shape
(cX, cY) = (int(w / 2.0), int(h / 2.0))
# compute the FFT to find the frequency transform, then shift
# the zero frequency component (i.e., DC component located at
# the top-left corner) to the center where it will be more
# easy to analyze
fft = np.fft.fft2(image)
fftShift = np.fft.fftshift(fft)
# check to see if we are visualizing our output
if vis:
# compute the magnitude spectrum of the transform
magnitude = 20 * np.log(np.abs(fftShift))
# display the original input image
(fig, ax) = plt.subplots(1, 2, )
ax[0].imshow(image, cmap="gray")
ax[0].set_title("Input")
ax[0].set_xticks([])
ax[0].set_yticks([])
# display the magnitude image
ax[1].imshow(magnitude, cmap="gray")
ax[1].set_title("Magnitude Spectrum")
ax[1].set_xticks([])
ax[1].set_yticks([])
# show our plots
plt.show()
# zero-out the center of the FFT shift (i.e., remove low
# frequencies), apply the inverse shift such that the DC
# component once again becomes the top-left, and then apply
# the inverse FFT
fftShift[cY - size:cY + size, cX - size:cX + size] = 0
fftShift = np.fft.ifftshift(fftShift)
recon = np.fft.ifft2(fftShift)
# compute the magnitude spectrum of the reconstructed image,
# then compute the mean of the magnitude values
magnitude = 20 * np.log(np.abs(recon))
mean = np.mean(magnitude)
# the image will be considered "blurry" if the mean value of the
# magnitudes is less than the threshold value
return (mean, mean <= thresh)
pathImg = "path to the image"
image = cv2.imread(pathImg)
# Resizing the image to 500 pixels in width.
image = imutils.resize(image, width= 500)
# Converting the image to gray scale.
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# Using the FFT to detect blur.
(mean, blurry) = detect_blur_fft(gray, size=60)
image = np.dstack([gray] * 3)
# This is a conditional statement that will set the color to red if the image is blurry or green
color = (0, 0, 255) if blurry else (0, 255, 0)
text = "Blurry ({:.4f})" if blurry else "Not Blurry ({:.4f})"
text = text.format(mean)
# Adding text to the image.
cv2.putText(image, text, (10, 25), cv2.FONT_HERSHEY_SIMPLEX, 0.7, color, 2)
print("[INFO] {}".format(text))
# show the output image
display(pilImg.fromarray(image))

Pinch/bulge distortion using Python OpenCV

I want to apply a pinch/bulge filter on an image using Python OpenCV. The result should be some kind of this example:
https://pixijs.io/pixi-filters/tools/screenshots/dist/bulge-pinch.gif
I've read the following stackoverflow post that should be the correct formula for the filter: Formulas for Barrel/Pincushion distortion
But I'm struggling to implement this in Python OpenCV.
I've read about maps to apply filter on an image: Distortion effect using OpenCv-python
As for my understanding, the code could look something like this:
import numpy as np
import cv2 as cv
f_img = 'example.jpg'
im_cv = cv.imread(f_img)
# grab the dimensions of the image
(h, w, _) = im_cv.shape
# set up the x and y maps as float32
flex_x = np.zeros((h, w), np.float32)
flex_y = np.zeros((h, w), np.float32)
# create map with the barrel pincushion distortion formula
for y in range(h):
for x in range(w):
flex_x[y, x] = APPLY FORMULA TO X
flex_y[y, x] = APPLY FORMULA TO Y
# do the remap this is where the magic happens
dst = cv.remap(im_cv, flex_x, flex_y, cv.INTER_LINEAR)
cv.imshow('src', im_cv)
cv.imshow('dst', dst)
cv.waitKey(0)
cv.destroyAllWindows()
Is this the correct way to achieve the distortion presented in the example image? Any help regarding useful ressources or preferably examples are much appreciated.
After familiarizing myself with the ImageMagick source code, I've found a way to apply the formula for distortion. With the help of the OpenCV remap function, this is a way to distort an image:
import numpy as np
import cv2 as cv
f_img = 'example.jpg'
im_cv = cv.imread(f_img)
# grab the dimensions of the image
(h, w, _) = im_cv.shape
# set up the x and y maps as float32
flex_x = np.zeros((h, w), np.float32)
flex_y = np.zeros((h, w), np.float32)
# create map with the barrel pincushion distortion formula
for y in range(h):
delta_y = scale_y * (y - center_y)
for x in range(w):
# determine if pixel is within an ellipse
delta_x = scale_x * (x - center_x)
distance = delta_x * delta_x + delta_y * delta_y
if distance >= (radius * radius):
flex_x[y, x] = x
flex_y[y, x] = y
else:
factor = 1.0
if distance > 0.0:
factor = math.pow(math.sin(math.pi * math.sqrt(distance) / radius / 2), -amount)
flex_x[y, x] = factor * delta_x / scale_x + center_x
flex_y[y, x] = factor * delta_y / scale_y + center_y
# do the remap this is where the magic happens
dst = cv.remap(im_cv, flex_x, flex_y, cv.INTER_LINEAR)
cv.imshow('src', im_cv)
cv.imshow('dst', dst)
cv.waitKey(0)
cv.destroyAllWindows()
This has the same effect as using the convert -implode function from ImageMagick.
You can do that using implode and explode options in Python Wand, which uses ImageMagick.
Input:
from wand.image import Image
import numpy as np
import cv2
with Image(filename='zelda1.jpg') as img:
img.virtual_pixel = 'black'
img.implode(0.5)
img.save(filename='zelda1_implode.jpg')
# convert to opencv/numpy array format
img_implode_opencv = np.array(img)
img_implode_opencv = cv2.cvtColor(img_implode_opencv, cv2.COLOR_RGB2BGR)
with Image(filename='zelda1.jpg') as img:
img.virtual_pixel = 'black'
img.implode(-0.5 )
img.save(filename='zelda1_explode.jpg')
# convert to opencv/numpy array format
img_explode_opencv = np.array(img)
img_explode_opencv = cv2.cvtColor(img_explode_opencv, cv2.COLOR_RGB2BGR)
# display result with opencv
cv2.imshow("IMPLODE", img_implode_opencv)
cv2.imshow("EXPLODE", img_explode_opencv)
cv2.waitKey(0)
Implode:
Explode:

Image Reconstruction (Erosion and Dilation) both gives black image as output

I am trying to remove the black spots from a face of this image using the erosion methods.
I have implemented:
img = skimage.io.imread('blemish.jpeg')
img = skimage.color.rgb2gray(img)
img_inten = skimage.exposure.rescale_intensity(img,in_range=(50,100))
diliation_seed = img_inten.copy()
diliation_seed[1:-1,1:-1] = img_inten.min()
mask = img_inten
eroded_img = skimage.morphology.reconstruction(diliation_seed,mask,method='dilation')
matplotlib.pyplot.imshow(eroded_img,cmap='gray')
My output is always a black image in both the cases. What is going wrong here?
rgb2gray is outputting an image as a matrix of floats, with values in [0;1]
So the rescale_intensity is just outputting a matrix of 0, since you ask for values between 50 and 100 and there is none in the gray img.
you can fix it like this :
import skimage
from skimage import data, exposure, img_as_float
from skimage.morphology import reconstruction
import matplotlib.pyplot as plt
img = skimage.io.imread('blemish.jpeg')
gray_img = 255*skimage.color.rgb2gray(img) # multiply by 255 to get back in the [0;255] range
img_inten = exposure.rescale_intensity(gray_img,in_range=(50,100))
diliation_seed = img_inten.copy()
diliation_seed[1:-1,1:-1] = img_inten.min()
mask = img_inten
eroded_img = reconstruction(diliation_seed,mask,method='dilation')
plt.imshow(eroded_img,cmap='gray')
plt.show()

Find average location of pixel of a given color

I am making an automatic curve detection program (graphs, 2D plots, etc.) and have gotten stuck with trying to find the average location of a color. The color is determined based on a K means clustering, so I have the RGB value for that color. I have also built a small sliding window program to follow a set of (X,Y) coordinates, and in that window is where the average location/centroid for a particular color needs to be found. I am very new to OpenCV, so I will detail what I want to do below, and appreciate any help.
Find the 3 most common colors with K-means and obtain the RGB values of those colors (done).
Use the sliding window to establish an N x N search window that rides along a determined X,Y path (done).
Apply a mask to get the curve color (based on Kmeans) extracted in the search window (need help). Also potentially mask out the background color based on Kmeans.
Find the average X,Y coordinate/pixel location for the desired curve color in the NxN search window (need help).
It would be appreciated if anyone could help with masking the color of the image based on the RGB value in OpenCV and then determining the average pixel location of the desired color in the search window. I have included the code that I have so far. The boundaries would come from the K-means clustering, but right now they are just hard coded for testing purposes.
Below is an example curve, where I would attempt to follow any of the black lines
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import numpy as np
import cv2
import math
import time
from tkinter import *
from tkinter import filedialog
from tkinter import ttk
from tkinter.filedialog import askopenfilename
from tkinter import scrolledtext
from PIL import Image, ImageTk
import os
import ReadCoords
from sklearn.cluster import KMeans
(winW, winH) = (12,12)
xs,ys = ReadCoords.read_coords()
image = cv2.imread(r"Curve.PNG")
boundaries = [([0,0,0],[128,128,128])]
def color_detection(image, boundaries):
for (lower,upper) in boundaries:
lower = np.array(lower, dtype = 'uint8')
upper = np.array(upper, dtype = 'uint8')
mask = cv2.inRange(image, lower, upper)
output = cv2.bitwise_and(image, image, mask = mask)
return output
def kmeans_colors(image = image):
org_clone = image.copy()
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
# image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
image = image.reshape((image.shape[0] * image.shape[1], 3))
# image = image.reshape((image.shape[0] * image.shape[1], 1))
clt = KMeans(n_clusters = 3)
clt.fit(image)
# print(clt.cluster_centers_)
def centroid_histogram(clt):
# grab the number of different clusters and create a histogram
# based on the number of pixels assigned to each cluster
numLabels = np.arange(0, len(np.unique(clt.labels_)) + 1)
(hist, _) = np.histogram(clt.labels_, bins = numLabels)
# normalize the histogram, such that it sums to one
hist = hist.astype("float")
hist /= hist.sum()
# return the histogram
return hist
def plot_colors(hist, centroids):
# initialize the bar chart representing the relative frequency
# of each of the colors
bar = np.zeros((50, 300, 3), dtype = "uint8")
startX = 0
# loop over the percentage of each cluster and the color of
# each cluster
for (percent, color) in zip(hist, centroids):
# plot the relative percentage of each cluster
endX = startX + (percent * 300)
cv2.rectangle(bar, (int(startX), 0), (int(endX), 50),
color.astype("uint8").tolist(), -1)
startX = endX
# return the bar chart
return bar
hist = centroid_histogram(clt)
bar = plot_colors(hist, clt.cluster_centers_)
return clt.cluster_centers_
def curve_sliding_window(image, step_size, window_size, x_range, y_range):
for x,y in zip(x_range, y_range):
yield (x, y, image[y:y+window_size[1], x:x+window_size[0]])
for (x, y, window) in curve_sliding_window(image, step_size=8, window_size=(winW, winH), x_range = xs, y_range = ys ):
# if the window does not meet our desired window size, ignore it
if window.shape[0] != winH or window.shape[1] != winW:
continue
# THIS IS WHERE YOU WOULD PROCESS YOUR WINDOW, SUCH AS APPLYING A
# MACHINE LEARNING CLASSIFIER TO CLASSIFY THE CONTENTS OF THE
# WINDOW
# since we do not have a classifier, we'll just draw the window
clone = image.copy()
cv2.rectangle(clone, (x, y), (x + winW, y + winH), (0, 0, 255), 2)
cv2.imshow("Window", clone)
# cv2.waitKey(1)
# time.sleep(0.01)
masked = color_detection(window, boundaries)
cX,cY = moment_centriod(contours(window))
if cX != 3 and cY != 3:
cv2.circle(window, (cX, cY), 1, (0,0,255), 2)
cv2.imshow("windows", np.hstack([masked, window]))
cv2.waitKey(50)

Using an image as a mask for another image

I've asked this question already, but I've given it a go since. Here's my
code, with a doc string to show what I'm trying to do.
"""
w1_w2_filter_mask2.py
Use xbulge-mask.fits image file as a mask for W1 and W2 bands
and compute the median in patches where the image is not masked
"""
import matplotlib.pyplot as plt
import numpy as np
from astropy.io import fits
from scipy.ndimage.filters import generic_filter as gf
# Open data files
hdulist = fits.open('xbulge-w1.fits')
w1data = hdulist[0].data
hdulist2 = fits.open('xbulge-w2.fits')
w2data = hdulist2[0].data
hdulistmask = fits.open('xbulge-mask.fits')
maskdata = hdulistmask[0].data
# Define physical shape of filter patch
def patch_filter(image_data, radius):
kernel = np.zeros((2*radius+1, 2*radius+1))
y, x = np.ogrid[-radius:radius+1, -radius:radius+1]
patch = x**2 + y**2 <= radius**2
kernel[patch] = 1
filtered_image = gf(image_data, np.median, footprint = kernel)
return filtered_image
# Apply mask to image files
mx1 = np.ma.masked_array(w1data, mask=maskdata)
mx2 = np.ma.masked_array(w2data, mask=maskdata)
# Pass median filtering patch across masked image
radius = 25
w1masked_filtered = patch_filter(mx1, radius)
w2masked_filtered = patch_filter(mx2, radius)
When I print or imshow the masked, filtered array, it has the same effect as when I don't apply the mask image, i.e. (load mask -> load image -> apply mask -> filter in patches) has the same effect as (load image -> filter in patches).
Can anyone see what I'm missing?

Categories