Dominant Color Extraction

#!/usr/bin/env python
# coding: utf-8
# http://www.pillalamarri.in/python/dominant-color-extraction/
# In[1]:


import numpy as np
import matplotlib.pyplot as plt
import cv2


# In[2]:


img= cv2.imread("images/1-Saint-Basils-Cathedral.jpg")


# In[3]:


img.shape


# In[4]:


plt.imshow(img)
plt.show()


# In[5]:


img.shape

all_pixels=img.reshape((-1,3))

all_pixels.shape

all_pixels.shape


# # Visualizing The Image through K-Means sklearn

# In[6]:


from sklearn.cluster import KMeans


# In[7]:


k=9
km=KMeans(n_clusters=k)


# In[8]:


km.fit(all_pixels)


# In[9]:


km.cluster_centers_


# In[10]:


centers=np.array(km.cluster_centers_,dtype="uint")
#uint8 to represent number btw 0-255 8 bit integer is sufficient 


# In[11]:


centers 


# ### plot all these color ( data centres)

# In[12]:


for c in centers:
    a=np.zeros((100,100,3))
    a[:,:,:]=c
    
    plt.imshow(a/255)
    plt.axis("off")
    plt.show()


# In[13]:


km.labels_


# In[14]:


new_img = np.zeros((all_pixels.shape[0], 3), dtype="uint")


# In[15]:


new_img.shape


# In[16]:


# giving appropriate colors to the pixel/data points of new image
for ix in range(all_pixels.shape[0]): #pic colors from centre and map it according to labels
    new_img[ix] = centers[km.labels_[ix]]


# or we can do it like this by using plt.figure and creating list of all colors

# In[17]:


i= 1

plt.figure(0,figsize=(10,2)) #fig size


colors = []

for each_col in centers:
    plt.subplot(1,len(centers),i) # 1 row 9 col
    plt.axis("off")
    i+=1 #everytime we go to next box we increment it!
    
    colors.append(each_col)
    
    #Color Swatch
    a = np.zeros((100,100,3),dtype='uint')
    a[:,:,:] = each_col
    plt.imshow(a)
    
plt.show()


for ix in range(new_img.shape[0]):
    new_img[ix] = colors[km.labels_[ix]]
    
new_img = new_img.reshape((img.shape))
plt.imshow(new_img)
plt.show()


# In[18]:


new_img = new_img.reshape((600,394,3))


# In[19]:


plt.imshow(new_img)
plt.show()


# In[20]:


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):
            #The zip() function returns an 
            #iterator of tuples based on the iterable objects
            # plot the relative percentage of each cluster
            #zip() returns an iterator of 
            #tuples with each tuple having only one element.
            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


# In[21]:


numLabels = np.arange(0, len(np.unique(km.labels_)) + 1)

(hist, _) = np.histogram(km.labels_, bins = numLabels)
# normalize the histogram, such that it sums to one
hist = hist.astype("float")
hist /= hist.sum()
       


# In[22]:


bar = plot_colors(hist, km.cluster_centers_)
# show our color bart
plt.imshow(img)
plt.axis("off")
plt.imshow(bar)
plt.show()


# In[23]:



img=cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
plt.imshow(img)
plt.show()


# In[24]:


img.shape


# In[25]:


all_pixels=img.reshape((-1,3)) 


# In[26]:


from sklearn.cluster import KMeans
k=9
km=KMeans(n_clusters=k)
#so it will take color from cluster and find color 
km.fit(all_pixels)
km.cluster_centers_ #center we have in decimal 
#so rgb(14.91798497,  17.94468754,  13.52473915)<== are some color as these ar some dominant points


# In[27]:


centers=np.array(km.cluster_centers_,dtype="uint") #so changing it in integers
#centers==>color
centers


# In[28]:


for c in centers:
    a=np.zeros((100,100,3)) #image 
    a[:,:,:]=c #so image will broadcast into a of the cluster center from img
    
    plt.imshow(a/255) #a should be in 0 and 1
    plt.axis("off") #removing axis 
    plt.show()


# In[29]:


km.labels_#it will give cluster number associated with a cluster
km.labels_.shape #showing number of cluster 


# In[30]:


new_img = np.zeros((all_pixels.shape[0], 3), dtype="uint")
new_img


# In[31]:


new_img.shape
#you have same shape of image eariler of Pixels


# In[32]:


# giving appropriate colors to the pixel/data points of new image
for ix in range(all_pixels.shape[0]):
    new_img[ix] = centers[km.labels_[ix]]
#in this we are iterating over all the points of image alloting all the color we have to new image
#colors are in center 


# In[37]:


new_img = new_img.reshape((600, 394,3))
plt.imshow(img)
plt.axis('off')
plt.show()
plt.axis('off')
plt.imshow(new_img)
plt.show()


# In[ ]:





# In[ ]:

http://www.pillalamarri.in/python/dominant-color-extraction/