Develop Reference

FFT on image with Python

I have a problem with FFT implementation in Python. I have completely strange results.
Ok so, I want to open image, get value of every pixel in RGB, then I need to use fft on it, and convert to image again.
My steps:
1) I'm opening image with PIL library in Python like this
from PIL import Image
im = Image.open("test.png")
2) I'm getting pixels
pixels = list(im.getdata())
3) I'm seperate every pixel to r,g,b values
for x in range(width):
for y in range(height):
r,g,b = pixels[x*width+y]
red[x][y] = r
green[x][y] = g
blue[x][y] = b
4). Let's assume that I have one pixel (111,111,111). And use fft on all red values like this
red = np.fft.fft(red)
And then:
print (red[0][0], green[0][0], blue[0][0])
My output is:
(53866+0j) 111 111
It's completely wrong I think. My image is 64x64, and FFT from gimp is completely different. Actually, my FFT give me only arrays with huge values, thats why my output image is black.
Do you have any idea where is problem?
[EDIT]
I've changed as suggested to
red= np.fft.fft2(red)
And after that I scale it
scale = 1/(width*height)
red= abs(red* scale)
And still, I'm getting only black image.
[EDIT2]
Ok, so lets take one image.
Assume that I dont want to open it and save as greyscale image. So I'm doing like this.
def getGray(pixel):
r,g,b = pixel
return (r+g+b)/3
im = Image.open("test.png")
im.load()
pixels = list(im.getdata())
width, height = im.size
for x in range(width):
for y in range(height):
greyscale[x][y] = getGray(pixels[x*width+y])
data = []
for x in range(width):
for y in range(height):
pix = greyscale[x][y]
data.append(pix)
img = Image.new("L", (width,height), "white")
img.putdata(data)
img.save('out.png')
After this, I'm getting this image , which is ok. So now, I want to make fft on my image before I'll save it to new one, so I'm doing like this
scale = 1/(width*height)
greyscale = np.fft.fft2(greyscale)
greyscale = abs(greyscale * scale)
after loading it. After saving it to file, I have . So lets try now open test.png with gimp and use FFT filter plugin. I'm getting this image, which is correct
How I can handle it?

Great question. I’ve never heard of it but the Gimp Fourier plugin seems really neat:
A simple plug-in to do fourier transform on you image. The major advantage of this plugin is to be able to work with the transformed image inside GIMP. You can so draw or apply filters in fourier space, and get the modified image with an inverse FFT.
This idea—of doing Gimp-style manipulation on frequency-domain data and transforming back to an image—is very cool! Despite years of working with FFTs, I’ve never thought about doing this. Instead of messing with Gimp plugins and C executables and ugliness, let’s do this in Python!
Caveat. I experimented with a number of ways to do this, attempting to get something close to the output Gimp Fourier image (gray with moiré pattern) from the original input image, but I simply couldn’t. The Gimp image appears to be somewhat symmetric around the middle of the image, but it’s not flipped vertically or horizontally, nor is it transpose-symmetric. I’d expect the plugin to be using a real 2D FFT to transform an H×W image into a H×W array of real-valued data in the frequency domain, in which case there would be no symmetry (it’s just the to-complex FFT that’s conjugate-symmetric for real-valued inputs like images). So I gave up trying to reverse-engineer what the Gimp plugin is doing and looked at how I’d do this from scratch.
The code. Very simple: read an image, apply scipy.fftpack.rfft in the leading two dimensions to get the “frequency-image”, rescale to 0–255, and save.
Note how this is different from the other answers! No grayscaling—the 2D real-to-real FFT happens independently on all three channels. No abs needed: the frequency-domain image can legitimately have negative values, and if you make them positive, you can’t recover your original image. (Also a nice feature: no compromises on image size. The size of the array remains the same before and after the FFT, whether the width/height is even or odd.)
from PIL import Image
import numpy as np
import scipy.fftpack as fp
## Functions to go from image to frequency-image and back
im2freq = lambda data: fp.rfft(fp.rfft(data, axis=0),
axis=1)
freq2im = lambda f: fp.irfft(fp.irfft(f, axis=1),
axis=0)
## Read in data file and transform
data = np.array(Image.open('test.png'))
freq = im2freq(data)
back = freq2im(freq)
# Make sure the forward and backward transforms work!
assert(np.allclose(data, back))
## Helper functions to rescale a frequency-image to [0, 255] and save
remmax = lambda x: x/x.max()
remmin = lambda x: x - np.amin(x, axis=(0,1), keepdims=True)
touint8 = lambda x: (remmax(remmin(x))*(256-1e-4)).astype(int)
def arr2im(data, fname):
out = Image.new('RGB', data.shape[1::-1])
out.putdata(map(tuple, data.reshape(-1, 3)))
out.save(fname)
arr2im(touint8(freq), 'freq.png')
(Aside: FFT-lover geek note. Look at the documentation for rfft for details, but I used Scipy’s FFTPACK module because its rfft interleaves real and imaginary components of a single pixel as two adjacent real values, guaranteeing that the output for any-sized 2D image (even vs odd, width vs height) will be preserved. This is in contrast to Numpy’s numpy.fft.rfft2 which, because it returns complex data of size width/2+1 by height/2+1, forces you to deal with one extra row/column and deal with deinterleaving complex-to-real yourself. Who needs that hassle for this application.)
Results. Given input named test.png:
this snippet produces the following output (global min/max have been rescaled and quantized to 0-255):
And upscaled:
In this frequency-image, the DC (0 Hz frequency) component is in the top-left, and frequencies move higher as you go right and down.
Now, let’s see what happens when you manipulate this image in a couple of ways. Instead of this test image, let’s use a cat photo.
I made a few mask images in Gimp that I then load into Python and multiply the frequency-image with to see what effect the mask has on the image.
Here’s the code:
# Make frequency-image of cat photo
freq = im2freq(np.array(Image.open('cat.jpg')))
# Load three frequency-domain masks (DSP "filters")
bpfMask = np.array(Image.open('cat-mask-bpfcorner.png')).astype(float) / 255
hpfMask = np.array(Image.open('cat-mask-hpfcorner.png')).astype(float) / 255
lpfMask = np.array(Image.open('cat-mask-corner.png')).astype(float) / 255
# Apply each filter and save the output
arr2im(touint8(freq2im(freq * bpfMask)), 'cat-bpf.png')
arr2im(touint8(freq2im(freq * hpfMask)), 'cat-hpf.png')
arr2im(touint8(freq2im(freq * lpfMask)), 'cat-lpf.png')
Here’s a low-pass filter mask on the left, and on the right, the result—click to see the full-res image:
In the mask, black = 0.0, white = 1.0. So the lowest frequencies are kept here (white), while the high ones are blocked (black). This blurs the image by attenuating high frequencies. Low-pass filters are used all over the place, including when decimating (“downsampling”) an image (though they will be shaped much more carefully than me drawing in Gimp 😜).
Here’s a band-pass filter, where the lowest frequencies (see that bit of white in the top-left corner?) and high frequencies are kept, but the middling-frequencies are blocked. Quite bizarre!
Here’s a high-pass filter, where the top-left corner that was left white in the above mask is blacked out:
This is how edge-detection works.
Postscript. Someone, make a webapp using this technique that lets you draw masks and apply them to an image real-time!!!

There are several issues here.
1) Manual conversion to grayscale isn't good. Use Image.open("test.png").convert('L')
2) Most likely there is an issue with types. You shouldn't pass np.ndarray from fft2 to a PIL image without being sure their types are compatible. abs(np.fft.fft2(something)) will return you an array of type np.float32 or something like this, whereas PIL image is going to receive something like an array of type np.uint8.
3) Scaling suggested in the comments looks wrong. You actually need your values to fit into 0..255 range.
Here's my code that addresses these 3 points:
import numpy as np
from PIL import Image
def fft(channel):
fft = np.fft.fft2(channel)
fft *= 255.0 / fft.max() # proper scaling into 0..255 range
return np.absolute(fft)
input_image = Image.open("test.png")
channels = input_image.split() # splits an image into R, G, B channels
result_array = np.zeros_like(input_image) # make sure data types,
# sizes and numbers of channels of input and output numpy arrays are the save
if len(channels) > 1: # grayscale images have only one channel
for i, channel in enumerate(channels):
result_array[..., i] = fft(channel)
else:
result_array[...] = fft(channels[0])
result_image = Image.fromarray(result_array)
result_image.save('out.png')
I must admit I haven't managed to get results identical to the GIMP FFT plugin. As far as I see it does some post-processing. My results are all kinda very low contrast mess, and GIMP seems to overcome this by tuning contrast and scaling down non-informative channels (in your case all chanels except Red are just empty). Refer to the image:

How to detect a laser line in an image using Python

What's the quickest most reliable method of detecting a roughly horizontal red laser line in an image using Python? I'm working on a small project related to 3d laser scanning, and I need to be able to detect the laser in an image in order to calculate distance from its distortion.
To start, I have two images, a reference image A known to contain no laser line, and an image B that definitely contains a laser line, possibly distorted. e.g.
Sample image A:
Sample image B:
Since these are RGB, but the laser is red, I remove some noise by stripping out the blue and green channels using this function:
from PIL import Image
import numpy as np
def only_red(im):
"""
Strips out everything except red.
"""
data = np.array(im)
red, green, blue, alpha = data.T
im2 = Image.fromarray(red.T)
return im2
That gets me these images:
Next, I try and eliminate more noise by taking the difference of these two images using PIL.ImageChops.difference(). Ideally, the exposure between the two images would be identical, causing the difference to contain nothing except the laser line. Unfortunately, because the laser adds light, the exposure and overall brightness of each image is significantly different, resulting in a difference that still has considerable noise. e.g.
My final step is to make a "best guess" as to where the line is. Since I know the line will be roughly horizontal and the laser line should be the brightest thing in the image, I scan each column and find the row with the brightest pixel, which I assume to be the laser line. The code for this is:
import os
from PIL import Image, ImageOps
import numpy as np
x = Image.open('laser-diff.png', 'r')
x = x.convert('L')
out = Image.new("L", x.size, "black")
pix = out.load()
y = np.asarray(x.getdata(), dtype=np.float64).reshape((x.size[1], x.size[0]))
print y.shape
for col_i in xrange(y.shape[1]):
col_max = max([(y[row_i][col_i], row_i) for row_i in xrange(y.shape[0])])
col_max_brightness, col_max_row = col_max
print col_i, col_max
pix[col_i, col_max_row] = 255
out.save('laser-line.png')
All I really need to perform my distance calculation is the array of col_max values, but the laser-line.png helps me visualize the success, and looks like:
As you can see, the estimate is pretty close, but it still has some noise, mostly on the left-hand side of the image where the laser line is absorbed by a matte black finish.
What can I do to improve my accuracy and/or speed? I'm trying to run this on an ARM platform like the Raspberry Pi, so I'm worried my code might to be too inefficient to run well.
I'm not fully familiar with Numpy's matrix functions, so I had to settle for a slow for loop to scan each column instead of something more efficient. Is there a fast way to find the row with the brightest pixel per column in Numpy?
Also, is there a reliable way to equalize the images prior to performing the difference without dimming the laser line?

First enter the color that is the laser and leaves only the red color (in this case). Then apply the same effects and check the result.
In this case, you will have a much less polluted result.
Result
A problem is encountered in analyzing the red on the door, that has been lost.

I tried to do something. I don't think it's totally robust. But on your example it works relatively well.
I used canny edge detection to detect edge in your "difference" image. And then applied the Hough line transform as in this tutorial.
So I started with your processed image (that I call lineDetection.jpg in the code).
Here is the final script
import cv2
import numpy as np
img = cv2.imread('lineDetection.jpg')
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(gray,10,100)
minLineLength = 50
maxLineGap = 20
lines = cv2.HoughLinesP(edges,0.05,np.pi/5000,10,minLineLength,maxLineGap)
print(len(lines))
for i in range(len(lines)):
x1,y1,x2,y2 = lines[i][0]
cv2.line(img,(x1,y1),(x2,y2),(0,255,0),2)
cv2.imwrite('houghlines5.jpg',img)
In green line detected on the processed image. (You could add it to the original image for nicer effect)
Hope it helps.

First you can probably rescale the intensity of your negative image before subtracting it from your positive, to remove more noise. For example maybe rescaling by the ratios of the average intesity might be a good first try?
You can also try to put a threshold: if your max in below whatever good value, then it is probably not your laser but a noisy point...
Then yes numpy can find the best row / col with the argmax function.

Edge detection for image stored in matrix

I represent images in the form of 2-D arrays. I have this picture:
How can I get the pixels that are directly on the boundaries of the gray region and colorize them?
I want to get the coordinates of the matrix elements in green and red separately. I have only white, black and gray regions on the matrix.

The following should hopefully be okay for your needs (or at least help). The idea is to split into the various regions using logical checks based on threshold values. The edge between these regions can then be detected using numpy roll to shift pixels in x and y and comparing to see if we are at an edge,
import matplotlib.pyplot as plt
import numpy as np
import scipy as sp
from skimage.morphology import closing
thresh1 = 127
thresh2 = 254
#Load image
im = sp.misc.imread('jBD9j.png')
#Get threashold mask for different regions
gryim = np.mean(im[:,:,0:2],2)
region1 = (thresh1<gryim)
region2 = (thresh2<gryim)
nregion1 = ~ region1
nregion2 = ~ region2
#Plot figure and two regions
fig, axs = plt.subplots(2,2)
axs[0,0].imshow(im)
axs[0,1].imshow(region1)
axs[1,0].imshow(region2)
#Clean up any holes, etc (not needed for simple figures here)
#region1 = sp.ndimage.morphology.binary_closing(region1)
#region1 = sp.ndimage.morphology.binary_fill_holes(region1)
#region1.astype('bool')
#region2 = sp.ndimage.morphology.binary_closing(region2)
#region2 = sp.ndimage.morphology.binary_fill_holes(region2)
#region2.astype('bool')
#Get location of edge by comparing array to it's
#inverse shifted by a few pixels
shift = -2
edgex1 = (region1 ^ np.roll(nregion1,shift=shift,axis=0))
edgey1 = (region1 ^ np.roll(nregion1,shift=shift,axis=1))
edgex2 = (region2 ^ np.roll(nregion2,shift=shift,axis=0))
edgey2 = (region2 ^ np.roll(nregion2,shift=shift,axis=1))
#Plot location of edge over image
axs[1,1].imshow(im)
axs[1,1].contour(edgex1,2,colors='r',lw=2.)
axs[1,1].contour(edgey1,2,colors='r',lw=2.)
axs[1,1].contour(edgex2,2,colors='g',lw=2.)
axs[1,1].contour(edgey2,2,colors='g',lw=2.)
plt.show()
Which gives the . For simplicity I've use roll with the inverse of each region. You could roll each successive region onto the next to detect edges
Thank you to #Kabyle for offering a reward, this is a problem that I spent a while looking for a solution to. I tried scipy skeletonize, feature.canny, topology module and openCV with limited success... This way was the most robust for my case (droplet interface tracking). Hope it helps!

There is a very simple solution to this: by definition any pixel which has both white and gray neighbors is on your "red" edge, and gray and black neighbors is on the "green" edge. The lightest/darkest neighbors are returned by the maximum/minimum filters in skimage.filters.rank, and a binary combination of masks of pixels that have a lightest/darkest neighbor which is white/gray or gray/black respectively produce the edges.
Result:
A worked solution:
import numpy
import skimage.filters.rank
import skimage.morphology
import skimage.io
# convert image to a uint8 image which only has 0, 128 and 255 values
# the source png image provided has other levels in it so it needs to be thresholded - adjust the thresholding method for your data
img_raw = skimage.io.imread('jBD9j.png', as_grey=True)
img = numpy.zeros_like(img, dtype=numpy.uint8)
img[:,:] = 128
img[ img_raw < 0.25 ] = 0
img[ img_raw > 0.75 ] = 255
# define "next to" - this may be a square, diamond, etc
selem = skimage.morphology.disk(1)
# create masks for the two kinds of edges
black_gray_edges = (skimage.filters.rank.minimum(img, selem) == 0) & (skimage.filters.rank.maximum(img, selem) == 128)
gray_white_edges = (skimage.filters.rank.minimum(img, selem) == 128) & (skimage.filters.rank.maximum(img, selem) == 255)
# create a color image
img_result = numpy.dstack( [img,img,img] )
# assign colors to edge masks
img_result[ black_gray_edges, : ] = numpy.asarray( [ 0, 255, 0 ] )
img_result[ gray_white_edges, : ] = numpy.asarray( [ 255, 0, 0 ] )
imshow(img_result)
P.S. Pixels which have black and white neighbors, or all three colors neighbors, are in an undefined category. The code above doesn't color those. You need to figure out how you want the output to be colored in those cases; but it is easy to extend the approach above to produce another mask or two for that.
P.S. The edges are two pixels wide. There is no getting around that without more information: the edges are between two areas, and you haven't defined which one of the two areas you want them to overlap in each case, so the only symmetrical solution is to overlap both areas by one pixel.
P.S. This counts the pixel itself as its own neighbor. An isolated white or black pixel on gray, or vice versa, will be considered as an edge (as well as all the pixels around it).

While plonser's answer may be rather straight forward to implement, I see it failing when it comes to sharp and thin edges. Nevertheless, I suggest you use part of his approach as preconditioning.
In a second step you want to use the Marching Squares Algorithm. According to the documentation of scikit-image, it is
a special case of the marching cubes algorithm (Lorensen, William and
Harvey E. Cline. Marching Cubes: A High Resolution 3D Surface
Construction Algorithm. Computer Graphics (SIGGRAPH 87 Proceedings)
21(4) July 1987, p. 163-170
There even exists a Python implementation as part of the scikit-image package. I have been using this algorithm (my own Fortran implementation, though) successfully for edge detection of eye diagrams in communications engineering.
Ad 1: Preconditioning
Create a copy of your image and make it two color only, e.g. black/white. The coordinates remain the same, but you make sure that the algorithm can properly make a yes/no-decision independent from the values that you use in your matrix representation of the image.
Ad 2: Edge Detection
Wikipedia as well as various blogs provide you with a pretty elaborate description of the algorithm in various languages, so I will not go into it's details. However, let me give you some practical advice:
Your image has open boundaries at the bottom. Instead of modifying the algorithm, you can artifically add another row of pixels (black or grey to bound the white/grey areas).
The choice of the starting point is critical. If there are not too many images to be processed, I suggest you select it manually. Otherwise you will need to define rules. Since the Marching Squares Algorithm can start anywhere inside a bounded area, you could choose any pixel of a given color/value to detect the corresponding edge (it will initially start walking in one direction to find an edge).
The algorithm returns the exact 2D positions, e.g. (x/y)-tuples. You can either
iterate through the list and colorize the corresponding pixels by assigning a different value or
create a mask to select parts of your matrix and assign the value that corresponds to a different color, e.g. green or red.
Finally: Some Post-Processing
I suggested to add an artificial boundary to the image. This has two advantages:
1. The Marching Squares Algorithm works out of the box.
2. There is no need to distinguish between image boundary and the interface between two areas within the image. Just remove the artificial boundary once you are done setting the colorful edges -- this will remove the colored lines at the boundary of the image.

Basically by follow pyStarter's suggestion of using the marching square algorithm from scikit-image, the desired could contours can be extracted with the following code:
import matplotlib.pyplot as plt
import numpy as np
import scipy as sp
from skimage import measure
import scipy.ndimage as ndimage
from skimage.color import rgb2gray
from pprint import pprint
#Load image
im = rgb2gray(sp.misc.imread('jBD9j.png'))
n, bins_edges = np.histogram(im.flatten(),bins = 100)
# Skip the black area, and assume two distinct regions, white and grey
max_counts = np.sort(n[bins_edges[0:-1] > 0])[-2:]
thresholds = np.select(
[max_counts[i] == n for i in range(max_counts.shape[0])],
[bins_edges[0:-1]] * max_counts.shape[0]
)
# filter our the non zero values
thresholds = thresholds[thresholds > 0]
fig, axs = plt.subplots()
# Display image
axs.imshow(im, interpolation='nearest', cmap=plt.cm.gray)
colors = ['r','g']
for i, threshold in enumerate(thresholds):
contours = measure.find_contours(im, threshold)
# Display all contours found for this threshold
for n, contour in enumerate(contours):
axs.plot(contour[:,1], contour[:,0],colors[i], lw = 4)
axs.axis('image')
axs.set_xticks([])
axs.set_yticks([])
plt.show()
!
However, from your image there is no clear defined gray region, so I took the two largest counts of intensities in the image and thresholded on these. A bit disturbing is the red region in the middle of the white region, however I think this could be tweaked with the number of bins in the histogram procedure. You could also set these manually as Ed Smith did.

Maybe there is a more elegant way to do that ...
but in case your array is a numpy array with dimensions (N,N) (gray scale) you can do
import numpy as np
# assuming black -> 0 and white -> 1 and grey -> 0.5
black_reg = np.where(a < 0.1, a, 10)
white_reg = np.where(a > 0.9, a, 10)
xx_black,yy_black = np.gradient(black_reg)
xx_white,yy_white = np.gradient(white_reg)
# getting the coordinates
coord_green = np.argwhere(xx_black**2 + yy_black**2>0.2)
coord_red = np.argwhere(xx_white**2 + yy_white**2>0.2)
The number 0.2 is just a threshold and needs to be adjusted.

I think you are probably looking for edge detection method for gray scale images. There are many ways to do that. Maybe this can help http://en.m.wikipedia.org/wiki/Edge_detection. For differentiating edges between white and gray and edges between black and gray, try use local average intensity.

microscopy image segmentation: bacteria segmentation with python

I am trying to segment some microscopy bright-field images showing some E. coli bacteria.
The picture I am working with resembles this one (even if this one is obtained with phase contrast):
my problem is that after running my segmentation function (OtsuMask below) I cannot distinguish dividing bacteria (you can try my code below on the sample image). This means that I get one single labeled region for a couple of bacteria which are joined by their end, instead of two different labeled images.
The boundary between two dividing bacteria is too narrow to be highlighted by the morphological operations I perform on the thresholded image, but I guess there must be a way to achieve my goal.
Any ideas/suggestions?
import scipy as sp
import numpy as np
from scipy import optimize
import mahotas as mht
from scipy import ndimage
import pylab as plt
def OtsuMask(img,dilation_size=2,erosion_size=1,remove_size=500):
img_thres=np.asarray(img)
s=np.shape(img)
p0=np.array([0,0,0])
p0[0]=(img[0,0]-img[0,-1])/512.
p0[1]=(img[1,0]-img[1,-1])/512.
p0[2]=img.mean()
[x,y]=np.meshgrid(np.arange(s[1]),np.arange(s[0]))
p=fitplane(img,p0)
img=img-myplane(p,x,y)
m=img.min()
img=img-m
img=abs(img)
img=img.astype(uint16)
"""perform thresholding with Otsu"""
T = mht.thresholding.otsu(img,2)
print T
img_thres=img
img_thres[img<T*0.9]=0
img_thres[img>T*0.9]=1
img_thres=-img_thres+1
"""morphological operations"""
diskD=createDisk(dilation_size)
diskE=createDisk(erosion_size)
img_thres=ndimage.morphology.binary_dilation(img_thres,diskD)
labeled_im,N=mht.label(img_thres)
label_sizes=mht.labeled.labeled_size(labeled_im)
labeled_im=mht.labeled.remove_regions(labeled_im,np.where(label_sizes<remove_size))
figure();
imshow(labeled_im)
return labeled_im
def myplane(p,x,y):
return p[0]*x+p[1]*y+p[2]
def res(p,data,x,y):
a=(data-myplane(p,x,y));
return array(np.sum(np.abs(a**2)))
def fitplane(data,p0):
s=shape(data);
[x,y]=meshgrid(arange(s[1]),arange(s[0]));
print shape(x), shape(y)
p=optimize.fmin(res,p0,args=(data,x,y));
print p
return p
def createDisk( size ):
x, y = np.meshgrid( np.arange( -size, size ), np.arange( -size, size ) )
diskMask = ( ( x + .5 )**2 + ( y + .5 )**2 < size**2)
return diskMask
THE FIRST PART OF THE CODE IN OtsuMask CONSIST OF A PLANE FITTING AND SUBTRACTION.

A similar approach to the one described in this related stackoverflow answer can be used here.
It goes basically like this:
threshold your image, as you have done
apply a distance transform on the thresholded image
threshold the distance transform, so that only a small 'seed' part of each bacterium remains
label these seeds, giving each one a different shade of gray
(also add a labeled seed for the background)
execute the watershed algorithm with these seeds and the distance transformed image, to get the separatd contours of your bacteria
Check out the linked answer for some pictures that will make this much clearer.

A few thoughts:
Otsu may not be a good choice, as you may even use a fixed threshold (your bacteria are black).
Thresholding the image with any method will remove a lot of useful information.
I do not have a complete recipe for you, but even this very simple thing seems to give a lot of interesting information:
import matplotlib.pyplot as plt
import cv2
# cv2 is only used to read the image into an array, use only green channel
bact = cv.imread("/tmp/bacteria.png")[:,:,1]
# draw a contour image with fixed threshold 50
fig = plt.figure()
ax = fig.add_subplot(111)
ax.contourf(bact, levels=[0, 50], colors='k')
This gives:
This suggests that if you use contour-tracing techniques with fixed contours, you will receive quite nice-looking starting points for dilation and erosion. So, two differences in thresholding:
Contouring uses much more of the grayscale information than simple black/white thresholding.
The fixed threshold seems to work well with these images, and if illumination correction is needed, Otsu is not the best choice.

One day skimage Watershed segmentation was more useful for me, than any OpenCV samples. It uses some code borrowed from Cellprofiler project (python-based tool for sophisticated cell image analysis). Hint: use Euclidean distance transform from opencv, it's faster than scipy implementation. Also peak_local_max function has distance parameter, which useful for precise single cells distinguishing. I think this function is more robust in finding cell peaks than rude threshold (because intensity of cells may vary).
You can find scipy watershed implementation, but it has weird behavior.

python separate round particles by offsetting contours / shrinking polygones

I'm new to python and stuck..
I want to make a python script that allows me to separate adjacent particles on an image like this:
into separate regions like this:
I was suggested to use the watershed method, which as far as I understand it would give me a something like this:
EDIT Actually found out that this is distance transform and not watershed
Where I then could use a threshold to separate them.. Followed this openCV watershed guide but it only worked to cut out the particles. Was not able to "transform" the code to do what I want.
I then took another approach. Tried to use the openCV contours which gave me good contours of the particles. I have then been looking intensively for an easy way to perform polygon offset in order to shrink the edge like this:
Using the center from the offset contours (polygon) should give me the number of particles.. But I just haven been able to find a simple way to do edge offset / polygon shrinking with python.

Here is a script using numpy, scipy and the scikit-image (aka skimage). It makes use of local maxima extraction and watershading plus labeling (ie connected components extraction).
import numpy as np
import scipy.misc
import scipy.ndimage
import skimage.feature
import skimage.morphology
# parameters
THRESHOLD = 128
# read image
im = scipy.misc.imread("JPh65.png")
# convert to gray image
im = im.mean(axis=-1)
# find peaks
peak = skimage.feature.peak_local_max(im, threshold_rel=0.9, min_distance=10)
# make an image with peaks at 1
peak_im = np.zeros_like(im)
for p in peak:
peak_im[p[0], p[1]] = 1
# label peaks
peak_label, _ = scipy.ndimage.label(peak_im)
# propagate peak labels with watershed
labels = skimage.morphology.watershed(255 - im, peak_label)
# limit watershed labels to area where the image is intense enough
result = labels * (im > THRESHOLD)