main.py

import cv2
import matplotlib.pyplot as plt
import numpy as np
import scipy.ndimage.morphology as morphology
from skimage.morphology import disk
def binarize(img):
   #binarized using Otsu's Method after gaussian blurring
    blur = cv2.GaussianBlur(img,(5,5),0)
    ret,th = cv2.threshold(blur,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU) 
    #Image Inversion
    inv=255-th
    return inv
#image segmentation

def imgsegments(inv):
    kernel = np.ones((1,50),np.uint8)
    erosion = cv2.erode(inv,kernel,iterations = 1)
    colvector=cv2.reduce(erosion,1, cv2.REDUCE_AVG).reshape(-1)
    m = inv.shape[0]
    w = inv.shape[1]
    result = np.zeros((m,w))

    # Draw a line for each row
    for row in range(m):
        if colvector[row]>0:
            cv2.line(result, (0,row), (w,row), (255), 1)
    result.astype(np.uint8)
    kernel = np.ones((40,1),np.uint8)
    vererosion = cv2.erode(255-result,kernel,iterations = 1)
    vererosion=255-vererosion
    vererosion=vererosion.astype(np.uint8)
    ret,vererosion=cv2.threshold(vererosion,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
    hist = cv2.reduce(vererosion,1, cv2.REDUCE_AVG).reshape(-1)

    lines=[]
    flag=0
    start=0
    H,W = img.shape[:2]
    for y in range(H-1):
        if hist[y]==255 or y==H-2:
            if flag==0:
                lines.append(start+int((y-start)/2))
                #print(y)
                flag=1
        elif hist[y]==0 and flag==1:
            flag=0
            start=y
            #print(start)
            
    hist = cv2.reduce(result,1, cv2.REDUCE_AVG).reshape(-1)

    staves=[]
    flag=0
    start=0
    H,W = img.shape[:2]
    for y in range(H-1):
        if hist[y]<1 and hist[y+1]>1:
            staves.append(y)
    onlynotes=cv2.subtract(inv,result.astype(np.uint8))
    kernel = np.ones((4,1),np.uint8)
    onlynotes = cv2.erode(255-onlynotes,kernel,iterations = 1)
    onlynotes=255-onlynotes
    return staves,lines,onlynotes,inv
img = cv2.imread('mary.jpg',0)
inv,th=binarize(img)
staves,lines,onlynotes,inv=imgsegments(inv)

#rest detection
def detectrest(crop_img,staff):    
    vertical_img = np.copy(crop_img)
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5,2))
    vertical_img = cv2.erode(vertical_img, kernel, iterations=1)
    vertical_img = cv2.dilate(vertical_img, kernel, iterations=1)
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (1,4))
    vertical_img = cv2.erode(vertical_img, kernel, iterations=1)
    vertical_img = cv2.dilate(vertical_img, kernel, iterations=1)
    # variance
    s = np.var(vertical_img,axis=0)
    y = np.array(s)
    if staff==1:
        y[:650] = 0
    elif staff==2:
        y[len(y)-150:]=0
    elif staff==10:
        y[:250]=0
        y[len(y)-150:]=0
    maximum = np.max(y)
    maxi = np.where(y == maximum)
    #print(maxi)
    ma = maxi[0]
    # pixel location for rest
    init = ma[0] - 10
    fin = ma[0] + 10
    # masking objects other than rest
    result=np.copy(crop_img)
    if maximum>7000:
        result[:,0:init] = 0 
        result[:,fin:]=0
    else:
        result[:,:]=0
    return result
#halfnote detection 
def halfnote(crop_img):
    im=crop_img
    im2=cv2.morphologyEx(im, cv2.MORPH_CLOSE, disk(3.5))

    im2=im2-im
    im2=cv2.morphologyEx(im2, cv2.MORPH_OPEN, disk(2.5))
    kernel1 = cv2.getStructuringElement(cv2.MORPH_RECT,(30,30))
    im2=cv2.morphologyEx(im2, cv2.MORPH_DILATE, disk(5))
    
    return im2
#quarternote detection
def quarternote(crop_img):
    im=crop_img
    im2=cv2.morphologyEx(im, cv2.MORPH_OPEN, disk(10))
    return im2

i=0
result = cv2.cvtColor(th,cv2.COLOR_GRAY2RGB)

#mapping with colors for individual notes
while(i<len(lines)-1):
    if i==0:
        rest=detectrest(onlynotes[lines[i]:lines[i+1],:],1)
        backtorgb=np.copy(result[lines[i]:lines[i+1]])
        backtorgb[rest>0,0]=255
        backtorgb[rest>0,1]=0
        backtorgb[rest>0,2]=0
        temp=np.copy(onlynotes[lines[i]:lines[i+1]])
        temp[:,:550]=0
        half=halfnote(temp)
        backtorgb[half>0,0]=0
        backtorgb[half>0,1]=255
        backtorgb[half>0,2]=0
        quarter=quarternote(inv[lines[i]:lines[i+1]])
        backtorgb[quarter>0,0]=0
        backtorgb[quarter>0,1]=0
        backtorgb[quarter>0,2]=255
        result[lines[i]:lines[i+1]]=np.copy(backtorgb)
    elif i>0:
        rest=detectrest(onlynotes[lines[i]:lines[i+1],:],10)
        backtorgb=np.copy(result[lines[i]:lines[i+1]])
        backtorgb[rest>0,0]=255
        backtorgb[rest>0,1]=0
        backtorgb[rest>0,2]=0
        temp=np.copy(onlynotes[lines[i]:lines[i+1]])
        temp[:,:250]=0
        half=halfnote(temp)
        backtorgb[half>0,0]=0
        backtorgb[half>0,1]=255
        backtorgb[half>0,2]=0
        quarter=quarternote(inv[lines[i]:lines[i+1]])
        backtorgb[quarter>0,0]=0
        backtorgb[quarter>0,1]=0
        backtorgb[quarter>0,2]=255
        result[lines[i]:lines[i+1]]=np.copy(backtorgb)
    i=i+1
plt.figure(figsize=(10,10))
plt.imshow(result)