# "Sharpen" an image by  multiplying every pixel by 2, and then subtracting
# the average value of the neighborhood from it. This procedure keeps the
# overall image intensity unchanged, but will enhance high frequency elements
# by subtracting a blurred (unsharp) version of the image from it.
#
#See slide number 22 from IrfanEssa-CP-02-5-Filtering.pdf

#
# Jay Summet 2024
#
#Python 3.10, OpenCV 4.5.4
#

import cv2
import numpy as np


#Load source / input image as grayscale, also works on color images...
imgIn = cv2.imread("oscar.jpg", cv2.IMREAD_GRAYSCALE)
cv2.imshow("Original", imgIn)


#Create the identity filter, but X 2
kernel = np.zeros( (9,9), np.float32)
kernel[4,4] = 2.0   #Identity, times two!
print(kernel)

#Create a box filter:
boxFilter = np.ones( (9,9), np.float32) / 81.0
#Try leaving out the /81 for fun....
print(boxFilter)

#Subtract the two:
kernel = kernel - boxFilter

print(kernel)


#Note that we are subject to overflow and underflow here...but 
# filter2D clips top and bottom ranges on the output when working with uint8 
# images. So this is mostly fine as long as you don't have a very bright 
# or very dark pixel surrounded by the opposite type. If this bothers
# you, work with floats and then convert back....

custom = cv2.filter2D(imgIn, -1, kernel)
cv2.imshow("Sharpen with unsharp mask", custom)


cv2.waitKey(0)
cv2.startWindowThread()