bendzov_image_algorytmus/minimal_energy.py
2021-09-06 06:28:09 +02:00

122 lines
3.5 KiB
Python

import math
from numpy import array as a
import numpy as np
from PIL import Image
im = Image.open("debug.bmp")
gray = im.convert('LA')
asnumpy_gray = np.asarray(gray)
image = np.asarray(im)
bnsdfcnsd = np.asarray(im)
sobel_0 = [
[-0.125, 0, 0.125],
[-0.25, 0, 0.25],
[-0.125, 0, 0.125]
]
sobel_1 = [
[-0.125, -0.25, -0.125],
[0, 0, 0],
[0.125, 0.25, 0.125]
]
edge_values = a([
a([0, 1, 0.5, 0.8]),
a([0.5, 0.2, 0.3, 0.4]),
a([1, 0.5, 0.6, 0.7]),
a([0.2, 0.1, 0.1, 0.1])
])
asnumpy_gray = a([a([pixel[0] for pixel in row]) for row in asnumpy_gray])
image = [row.tolist() for row in image]
def get_minimal_energy_map_vertical(edge_values_):
minimal_energy_map = np.zeros(shape=edge_values_.shape)
minimal_energy_map[-1] = edge_values_[-1]
for i in range(-2, -len(edge_values_) - 1, -1):
for j in range(len(edge_values_[0])):
if j == 0:
minimal_energy_map[i][j] = edge_values_[i][j] + min(minimal_energy_map[i + 1][j:j + 2])
elif j == len(edge_values_[0]) - 1:
minimal_energy_map[i][j] = edge_values_[i][j] + min(minimal_energy_map[i + 1][j - 1:j + 1])
else:
minimal_energy_map[i][j] = edge_values_[i][j] + min(minimal_energy_map[i + 1][j - 1:j + 2])
return minimal_energy_map
def seam_carve_vertical(image, minimal_energy_map):
seam = np.zeros(len(minimal_energy_map), dtype=np.int32)
seam[0] = (minimal_energy_map[0].tolist()).index(min(minimal_energy_map[0]))
for i in range(len(minimal_energy_map) - 1):
j = seam[i]
if j == 0:
sub_array = minimal_energy_map[i + 1][:2].tolist()
seam[i + 1] = sub_array.index(min(sub_array))
elif j == len(minimal_energy_map[0]) - 1:
sub_array = minimal_energy_map[i + 1][-2:].tolist()
seam[i + 1] = j + sub_array.index(min(sub_array)) - 1
else:
sub_array = minimal_energy_map[i + 1][j - 1:j + 2].tolist()
seam[i + 1] = j + sub_array.index(min(sub_array)) - 1
for i in range(len(seam)):
#image[i] = image[i][:seam[i]] + image[i][seam[i] + 1:]
image[i][seam[i]] = [255, 0, 0]
return a([a(row, dtype=np.uint8) for row in image], dtype=np.uint8)
def get_edges_values(input):
img1 = np.zeros(shape=asnumpy_gray.shape)
for y in range(1, len(img1) - 1):
for x in range(1, len(img1[y]) - 1):
inx0 = 0.
inx1 = 0.
for ax in range(-1, 2):
for b in range(-1, 2):
try:
inx0 += asnumpy_gray[y + ax][x + b] * sobel_0[ax + 1][b + 1]
inx1 += asnumpy_gray[y + ax][x + b] * sobel_1[ax + 1][b + 1]
except:
pass
img1[y][x] = math.sqrt(inx0 ** 2 + inx1 ** 2)
return img1
def dostuff(input):
for y in range(len(input)):
for x in range(len(input[y])):
input[y][x] = (((input[y][x] - -9635.0) * (255 - 0)) / (255 - -9635.0)) + 0
return input
for i in range(1):
print(i)
#tmp = get_minimal_energy_map_vertical(get_edges_values(asnumpy_gray))
tmp = get_edges_values(asnumpy_gray)
tmp = dostuff(tmp)
image = seam_carve_vertical(image, tmp)
out = Image.fromarray(tmp)
image = [row.tolist() for row in image]
gray = out.convert('LA')
asnumpy_gray = np.asarray(gray)
asnumpy_gray = a([a([pixel[0] for pixel in row]) for row in asnumpy_gray])
print("done!")
out.show()