consolidate all repos to one for archive
This commit is contained in:
201
semester_4/osnove_racunalniskega_vida/Naloga_3/main.py
Normal file
201
semester_4/osnove_racunalniskega_vida/Naloga_3/main.py
Normal file
@@ -0,0 +1,201 @@
|
||||
import numpy as np
|
||||
import cv2
|
||||
import random
|
||||
|
||||
def findClosestCentroids3(img, centroids):
|
||||
closestCentroi = np.zeros((img.shape[0],img.shape[1]), np.uint8)
|
||||
klusters = centroids.shape[0]
|
||||
for i in range(img.shape[0]):
|
||||
for j in range(img.shape[1]):
|
||||
dist = np.zeros(klusters, np.float32)
|
||||
for l in range(klusters):
|
||||
# calculate the distance between the pixel and the centroid
|
||||
dist[l] = np.sqrt(
|
||||
((img[i][j][0] - centroids[l][0]) ** 2)+
|
||||
((img[i][j][1] - centroids[l][1]) ** 2)+
|
||||
((img[i][j][2] - centroids[l][2]) ** 2))
|
||||
|
||||
|
||||
#assign the pixel to the cluster with the nearest centroid human readable
|
||||
#closestCentroi[i, j] = np.argmin(dist) * 255 / (klusters - 1)
|
||||
closestCentroi[i, j] = np.argmin(dist)
|
||||
|
||||
return closestCentroi
|
||||
|
||||
def calculateNewCenter3(img, centroids, closestCentroi):
|
||||
newCentroids = np.zeros(centroids.shape, np.float32)
|
||||
klusters = centroids.shape[0]
|
||||
centroidsCount = np.zeros(klusters, np.float32)
|
||||
for i in range(img.shape[0]):
|
||||
for j in range(img.shape[1]):
|
||||
newCentroids[closestCentroi[i][j]] += img[i][j]
|
||||
centroidsCount[closestCentroi[i][j]] += 1
|
||||
|
||||
for i in range(klusters):
|
||||
if centroidsCount[i] == 0:
|
||||
centroidsCount[i] = 1
|
||||
newCentroids[i] /= centroidsCount[i]
|
||||
|
||||
return newCentroids
|
||||
|
||||
def kmeans3(img, centroids, iterations):
|
||||
for i in range(iterations):
|
||||
print("Iteration: ", i)
|
||||
closestCentroi = findClosestCentroids3(img, centroids)
|
||||
# cv2.imshow('closestCentroi', closestCentroi)
|
||||
# cv2.waitKey(0)
|
||||
# cv2.destroyAllWindows()
|
||||
centroids = calculateNewCenter3(img, centroids, closestCentroi)
|
||||
|
||||
commpressedImg = np.zeros(img.shape, np.uint8)
|
||||
for i in range(img.shape[0]):
|
||||
for j in range(img.shape[1]):
|
||||
commpressedImg[i][j] = centroids[closestCentroi[i][j]]
|
||||
|
||||
return commpressedImg
|
||||
|
||||
def findClosestCentroids5(img, centroids, locations):
|
||||
closestCentroi = np.zeros((img.shape[0],img.shape[1]), np.uint8)
|
||||
klusters = centroids.shape[0]
|
||||
for i in range(img.shape[0]):
|
||||
for j in range(img.shape[1]):
|
||||
dist = np.zeros(klusters, np.float32)
|
||||
for l in range(klusters):
|
||||
# calculate the distance between the pixel and the centroid
|
||||
dist[l] = np.sqrt(
|
||||
((img[i][j][0] - centroids[l][0]) ** 2)+
|
||||
((img[i][j][1] - centroids[l][1]) ** 2)+
|
||||
((img[i][j][2] - centroids[l][2]) ** 2)+
|
||||
((i - locations[l][0]) ** 2)+
|
||||
((j - locations[l][1]) ** 2))
|
||||
|
||||
|
||||
# assign the pixel to the cluster with the nearest centroid human readable
|
||||
#closestCentroi[i, j] = np.argmin(dist) * 255 / (klusters - 1)
|
||||
closestCentroi[i, j] = np.argmin(dist)
|
||||
|
||||
return closestCentroi
|
||||
|
||||
def calculateNewCenter5(img, centroids, locations, closestCentroi):
|
||||
newCentroids = np.zeros(centroids.shape, np.float32)
|
||||
newLocations = np.zeros(locations.shape, np.float32)
|
||||
klusters = centroids.shape[0]
|
||||
centroidsCount = np.zeros(klusters, np.float32)
|
||||
for i in range(img.shape[0]):
|
||||
for j in range(img.shape[1]):
|
||||
newCentroids[closestCentroi[i][j]] += img[i][j]
|
||||
newLocations[closestCentroi[i][j]] += [i,j]
|
||||
centroidsCount[closestCentroi[i][j]] += 1
|
||||
|
||||
for i in range(klusters):
|
||||
if centroidsCount[i] == 0:
|
||||
centroidsCount[i] = 1
|
||||
newCentroids[i] /= centroidsCount[i]
|
||||
newLocations[i] /= centroidsCount[i]
|
||||
|
||||
return newCentroids, newLocations
|
||||
|
||||
def kmeans5(img, centroids, locations, iterations):
|
||||
for i in range(iterations):
|
||||
print("Iteration: ", i)
|
||||
closestCentroi = findClosestCentroids5(img, centroids, locations)
|
||||
centroids, locations = calculateNewCenter5(img, centroids, locations, closestCentroi)
|
||||
|
||||
commpressedImg = np.zeros(img.shape, np.uint8)
|
||||
for i in range(img.shape[0]):
|
||||
for j in range(img.shape[1]):
|
||||
commpressedImg[i][j] = centroids[closestCentroi[i][j]]
|
||||
|
||||
return commpressedImg
|
||||
|
||||
def kmeans(img, centroids, locations, iterations, dim):
|
||||
|
||||
if dim == 5:
|
||||
commpressedImg = kmeans5(img, centroids, locations, iterations)
|
||||
else:
|
||||
commpressedImg = kmeans3(img, centroids, iterations)
|
||||
|
||||
return commpressedImg
|
||||
|
||||
def click_event(event, x, y, flags, params):
|
||||
global points
|
||||
global locations
|
||||
if event == cv2.EVENT_LBUTTONDOWN:
|
||||
print(f'({y},{x})')
|
||||
points = np.append(points, [img[y,x]] , axis=0)
|
||||
locations = np.append(locations, [[y,x]] , axis=0)
|
||||
|
||||
def selectPointsM(img):
|
||||
cv2.namedWindow('Select color')
|
||||
cv2.setMouseCallback('Select color', click_event)
|
||||
|
||||
while True:
|
||||
cv2.imshow('Select color', img)
|
||||
if cv2.waitKey(1) & 0xFF == 27:
|
||||
break
|
||||
|
||||
cv2.destroyAllWindows()
|
||||
|
||||
T = 50
|
||||
def selectPointsR(img, stDimenzij):
|
||||
global points
|
||||
global locations
|
||||
global T
|
||||
numOfPoints = input('Vnesi stevilo tock:')
|
||||
|
||||
i = 0
|
||||
while i < int(numOfPoints):
|
||||
x = random.randint(0, img.shape[0]-1)
|
||||
y = random.randint(0, img.shape[1]-1)
|
||||
point = img[x,y]
|
||||
location = [x,y]
|
||||
|
||||
add = True
|
||||
|
||||
if stDimenzij == 3:
|
||||
for j in range(points.shape[0]):
|
||||
dist = np.sqrt(
|
||||
((points[j][0] - point[0]) ** 2)+
|
||||
((points[j][1] - point[1]) ** 2)+
|
||||
((points[j][2] - point[2]) ** 2))
|
||||
if dist < T:
|
||||
add = False
|
||||
break
|
||||
else:
|
||||
for j in range(points.shape[0]):
|
||||
dist = np.sqrt(
|
||||
((points[j][0] - point[0]) ** 2)+
|
||||
((points[j][1] - point[1]) ** 2)+
|
||||
((points[j][2] - point[2]) ** 2)+
|
||||
((locations[j][0] - location[0]) ** 2)+
|
||||
((locations[j][1] - location[1]) ** 2))
|
||||
if dist < T:
|
||||
add = False
|
||||
break
|
||||
|
||||
if add:
|
||||
points = np.append(points, [point] , axis=0)
|
||||
locations = np.append(locations, [location] , axis=0)
|
||||
i += 1
|
||||
|
||||
|
||||
# read the input image
|
||||
img = cv2.imread('veg.jpg' , 1)
|
||||
points = np.empty((0, 3), dtype=np.float32)
|
||||
locations = np.empty((0, 2), dtype=np.float32)
|
||||
|
||||
stDimenzij = input('Vnesi stevilo dimenzij 3 ali 5:')
|
||||
stIteracij = input('Vnesi stevilo iteracij:')
|
||||
selectType = input('Vnesi 1 za rocno izbiro, 2 za random izbiro:')
|
||||
|
||||
if selectType == '1':
|
||||
selectPointsM(img)
|
||||
else:
|
||||
selectPointsR(img, int(stDimenzij))
|
||||
|
||||
commpressedImg = kmeans(img, points, locations, int(stIteracij), int(stDimenzij))
|
||||
|
||||
cv2.imshow('Original image',img)
|
||||
cv2.imshow('Compressed image',commpressedImg)
|
||||
cv2.waitKey(0)
|
||||
cv2.destroyAllWindows()
|
||||
Reference in New Issue
Block a user