CODEWITHSUNDEEP
pythonpixelpython-imaging-libraryeuclidean-distance
Mikhaylova
Mikhaylova

Reputation: 137

Calculate Euclidean distance for each item in a list

I'm using Pillow with Python 3.3, and for each image I have, I want to analyse all pixels colors and calculate if each color is nearest to white or to black. I've tried with Euclidean distance with this formula:

>>>a,b=(127,255) #color value of pixel at position (1,1)
>>>whitedist=sqrt((a-255)**2 + (b-255)**2)
>>>print (whitedist)
128.0

Pixels have two values because I'm working with a greyscale image. The thing is that of course I can't do this for each pixel of a 128x128 image, so I tried to write my first "for" cycle ever:

colors=im.getdata() #all pixel values
for item in col:
    for item in range (0, 127):
    print ("B") #if the value is between 0 and 127 the color is nearer to black

this was a totally stupid attempt, python went into an infinite loop and I had to quit it. My problem is that I have a raw idea of what to do but I don't know python so well to translate my reasoning into code. I think my reasoning it's not completely wrong but I can't find a way to translate it into code.

Since list(im.getdata()) returns a list of tuples, one tuple for each pixel containing two values, what I think I should do is to apply the formula whitedist=sqrt((a-255)**2 + (b-255)**2) to each tuple and if the result of the formula is <128, tag the pixel as "dark", if >128 then tag the pixel as "light", then I need to count all dark and light pixels and see if the image has more dark or light pixels.

Is it totally wrong to do this with a for cycle?

Upvotes: 0

Views: 1625

Answers (2)

EmbedWise
EmbedWise

Reputation: 174

With getdata you retrieve all pixels, so if the image size is 128 x 128 this means 16384 pixels. These are stored as 16384 values in the variable colors as you mentioned.

You can just loop over the complete values with one loop:

for item in colors:
  whitedist=sqrt((item[0]-255)**2 + (item[1]-255)**2)
  if whitedist <= 128:
     darkPixels++
  else:
     lightPixels++

If you know how the colors are represented you can calculate the euclidean-distance as you mentioned. Note that you did not specify what todo if the value is 128, and how you want to store the tags. If it is only necessary to know how many pixels are dark or light you can use the example above.

Upvotes: 2

Irshad Bhat
Irshad Bhat

Reputation: 8709

Try this:

colors = im.getdata() #all pixel values
# initialize lists for dark and light pixels
darkList=[]   
lightList=[]
# set counter for dark and light pixels
dark = 0  
light = 0  
for item in colors:  # iterate over each tuple
    if sqrt((item[0]-255)**2 + (item[1]-255)**2) < 128:  # See if Euclidean distance is less than 128
        lightList.append(item)  # append pixel to light pixel list
        light+=1   # increment light pixel counter
    else:
        darkList.append(item)  # append pixel to dark pixel list
        dark+=1   # increment dark pixel counter

print("Total light pixels =",light)
print(lightList)
print("Total dark pixels =",dark)
print(darkList)

Upvotes: 1

Related Questions