functions.py

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


#user function to plot the images
def imgplot(image,title):
    plt.imshow(image)
    plt.title(title)
    plt.tick_params(left = False, right = False , labelleft = False ,labelbottom = False, bottom = False)
    return

def revised_coeff(num):
    h,w = np.shape(num)
    k = 3 #kernal
    image_result = np.zeros((h,w))
    number_new = np.pad(num,[1,1], mode = "constant")
    
    for i in range(h):
        for j in range(w):
            mat = number_new[i:i+k,j:j+k]
            image_result[i,j] = mat.max()
            
    return image_result



def consistency_verify(binary):
    h,w = np.shape(binary)
    k = 3
    imag = np.ones((h,w))
    new = np.pad(binary,[1,1],mode= "constant")
    
    #if 1s are more than 0s , make 1 or else 0
    for i in range(h):
        for j in range(w):
            mat = new[i:i+k,j:j+k]
            if (np.sum(mat)>=5):
                imag[i][j]  = 1
            else:
                imag[i][j] = 0
    #negation            
    imag = 1 - imag
    new2 = np.pad(imag,[1,1],mode= "constant")
    final = np.ones((h,w))
                    
    for i in range(h):
        for j in range(w):
            mat = new2[i:i+k,j:j+k]
            if (np.sum(mat)>=5):
                final[i][j]  = 1
            else:
                final[i][j] = 0
    #negation            
    final = 1 - final
    
    return final


#max selection
def max_selection(first,second,mode = 1):
    coef_first = revised_coeff(first)
    coef_second = revised_coeff(second)
    h,w = np.shape(coef_first)
    
    # Binary map 
    binary_map = np.zeros((h,w));
    
    #for max_max
    # Here 0 means first image and 1 means second image
    for i in range(h):
        for j in range(w):
            if (abs(coef_first[i][j]) > abs(coef_second[i][j])):  
                binary_map[i][j] = 0
            else:
                binary_map[i][j] = 1
                
    verified_map = consistency_verify(binary_map)
    
    h,w = np.shape(verified_map)
    for i in range(h):
        for j in range(w):
            if(verified_map[i][j]==1): #if 1 use image 2 pixel value else keep same as image 1
                first[i][j] = second[i][j]
    
    return first

#wavelet function to apply wavelet decomposition
def wavelet(image1,image2, mode = 1):
    [cA1,(cH1,cV1,cD1)] = pywt.dwt2(image1,wavelet= "coif5" ,mode = 'periodization') #taking wavelet transform of first image
    [cA2,(cH2,cV2,cD2)] = pywt.dwt2(image2,wavelet= "coif5" ,mode = 'periodization') # wavelet transfrom of second image
    
    cA = max_selection(cA1,cA2,mode)
    cH = max_selection(cH1,cH2,mode)
    cV = max_selection(cV1,cV2,mode)
    cD = max_selection(cD1,cD2,mode)
    
           
    Final =pywt.idwt2([cA,(cH,cV,cD)],wavelet="coif5",mode = 'periodization') #inverse wavelet transform
    Final = np.uint8(Final)
    return Final


#Fusion of images   
def fuse_images(img1,img2):
    r_fusion = wavelet(img1[:,:,0],img2[:,:,0])
    g_fusion = wavelet(img1[:,:,1],img2[:,:,1])
    b_fusion = wavelet(img1[:,:,2],img2[:,:,2])
    
    fused = np.uint8(np.dstack([r_fusion,g_fusion,b_fusion]))
    
    return fused