-
Notifications
You must be signed in to change notification settings - Fork 0
/
img_processing_webgui.py
247 lines (226 loc) · 12.6 KB
/
img_processing_webgui.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
import os
import cv2
import numpy as np
import streamlit as st
from natsort import natsorted
from skimage import filters
st.set_page_config(page_title="Image-Comparison Example", layout="centered")
st.title("Image Processing Functions")
# Let the user select the image they want to use
image = st.file_uploader("Upload an image...", type=["jpg", "png", "jpeg", "tif", "tiff"])
if image is None: st.info("Please upload an image file."); st.stop()
if image.size > 5000000: st.info("Please upload an image file smaller than 5MB."); st.stop()
image = cv2.imdecode(np.frombuffer(image.read(), np.uint8), 1)
# Delete any existing image.png file and altered.png file
try: os.remove("image.png")
except: pass
try: os.remove("altered.png")
except: pass
with st.form(key='my_form', clear_on_submit=False):
st.subheader("Settings")
# Add a toggle button for including the before/after image
before_after = st.checkbox("Before/After slider tool", value=True)
col1, col2 = st.columns(2)
top_crop = col1.number_input("Top Crop %", 0, 100, 0)
bottom_crop = col2.number_input("Bottom Crop %", 0, 100, 0)
col1, col2 = st.columns(2)
left_crop = col1.number_input("Left Crop", 0, 100, 0)
right_crop = col2.number_input("Right Crop", 0, 100, 0)
# Resize the image to a user selected size
resize = st.slider('Resize', 0, 100, 100)
# Give the user some options for altering the image with multiselect
img_processing_operations = natsorted(["AWB", "Blur", "Grayscale",
"Denoise", "Contrast", "Brightness",
"Saturation", "Sharpness", "CLAHE",
"AWB_custom", "Gamma", "Remove-Background",
"Tile-Plakakia","Sobel_Vertical","Sobel_Horizontal"])
options = st.multiselect("Alter the image...", img_processing_operations)
if "AWB_custom" in options:
# Perform custom white balance i.e. the user can select the factor for each channel
col1, col2 = st.columns(2)
with col1: awb_b = st.slider('AWB B', 0.0, 10.0, 3.0)
with col2: awb_g = st.slider('AWB G', 0.0, 10.0, 3.0)
if "Blur" in options:
# Blur the image using a Gaussian filter and a slider for the sigma value
sigmablur = st.slider('Blur Sigma', 0.0, 10.0, 3.0)
if "Denoise" in options:
col1, col2, col3 = st.columns(3)
with col1: diameter = st.slider('Denoise Diameter', 0, 10, 3)
with col2: sigmaColor = st.slider('Denoise Sigma Color', 0, 10, 3)
with col3: sigmaSpace = st.slider('Denoise Sigma Space', 0, 10, 3)
if "Sharpness" in options:
sigmaSharpness = st.slider('Sharpness Sigma', 0.0, 10.0, 3.0)
if "Brightness" in options:
brightness = st.slider('Brightness', 0.0, 10.0, 3.0)
if "Gamma" in options:
gamma = st.slider('Gamma', 0.0, 10.0, 3.0)
if "Saturation" in options:
saturation = st.slider('Saturation', 0.0, 10.0, 3.0)
if "CLAHE" in options:
col1, col2 = st.columns(2)
with col1:
clipLimit = st.slider('CLAHE Clip Limit', 0.0, 10.0, 3.0)
with col2:
tileGridSize = st.slider('CLAHE Tile Grid Size', 0, 10, 3)
if "Remove-Background" in options:
# Let the user select alpha matting foreground and background threshold values
col1, col2, col3 = st.columns(3)
with col1: alpha_matting = st.checkbox("Alpha Matting")
with col2: bg_threshold = st.slider('Background Threshold', 0, 255, 10)
with col3: fg_threshold = st.slider('Foreground Threshold', 0, 255, 240)
if "Tile-Plakakia" in options:
# Let the user select the tile_size and the step_size
col1, col2 = st.columns(2)
with col1: tile_size = st.slider('Tile Size', 0, 100, 50)
with col2: step_size = st.slider('Step Size', 0, 100, 50)
# Add a submit button
submit_button = st.form_submit_button(label='Initialize / Update Image')
# Explain each of the image processing functions
with st.expander("ℹ️ Glossary 📖"):
st.markdown("""
- **AWB** - Automatic White Balance (https://stackoverflow.com/questions/46390779/automatic-white-balancing-with-grayworld-assumption)
- **AWB_custom** - Custom White Balance
- **Blur** - Blur the image using a Gaussian filter (https://docs.opencv.org/4.x/d4/d13/tutorial_py_filtering.html)
- **Brightness** - Increase or decrease the image brightness (https://docs.opencv.org/3.4/d3/dc1/tutorial_basic_linear_transform.html)
- **CLAHE** - Contrast Limited Adaptive Histogram Equalization (https://docs.opencv.org/4.x/d5/daf/tutorial_py_histogram_equalization.html)
- **Contrast** - Increase or decrease the image contrast (https://docs.opencv.org/3.4/d3/dc1/tutorial_basic_linear_transform.html)
- **Denoise** - Denoise the image using a bilateral filter (https://scikit-image.org/docs/stable/auto_examples/filters/plot_denoise.html)
- **Gamma** - Gamma Correction (https://docs.opencv.org/3.4/d3/dc1/tutorial_basic_linear_transform.html)
- **Grayscale** - Convert the image to grayscale (https://docs.opencv.org/3.4/de/d25/imgproc_color_conversions.html)
- **Saturation** - Increase or decrease the image saturation (https://docs.opencv.org/3.4/d3/dc1/tutorial_basic_linear_transform.html)
- **Sharpness** - Increase or decrease the image sharpness (https://docs.opencv.org/3.4/d3/dc1/tutorial_basic_linear_transform.html)
- **Remove-Background** - Remove the background from the image (https://pypi.org/project/rembg/)
- **Sobel_Vertical** - Sobel Vertical Edge Detection (https://docs.opencv.org/3.4/d2/d2c/tutorial_sobel_derivatives.html)
- **Sobel_Horizontal** - Sobel Horizontal Edge Detection (https://docs.opencv.org/3.4/d2/d2c/tutorial_sobel_derivatives.html)
- **Tile-Plakakia** - Tile the image into smaller images (https://www.github.com/kalfasyan/plakakia)
""")
if submit_button:
# Crop the image
image = image[:, int(image.shape[1] * (left_crop / 100.0)):int(image.shape[1] * (1 - (right_crop / 100.0))), :]
# Crop the image
image = image[int(image.shape[0] * (top_crop / 100.0)):int(image.shape[0] * (1 - (bottom_crop / 100.0))), :, :]
image = cv2.resize(image, (0, 0), fx=(resize / 100.0), fy=(resize / 100.0))
cv2.imwrite("image.png", image)
if "AWB" in options:
image = cv2.cvtColor(image, cv2.COLOR_BGR2LAB)
avg_a = np.average(image[:, :, 1])
avg_b = np.average(image[:, :, 2])
for x in range(image.shape[0]):
for y in range(image.shape[1]):
l, a, b = image[x, y, :]
# fix for CV correction
l *= 100 / 255.0
image[x, y, 1] = a - ((avg_a - 128) * (l / 100.0) * 1.1)
image[x, y, 2] = b - ((avg_b - 128) * (l / 100.0) * 1.1)
image = cv2.cvtColor(image, cv2.COLOR_LAB2BGR)
# Explanation of the above code:
# We are using the "Gray World" assumption method to perform white balancing
# The idea is that in an image the average value for each channel is the same
# So we calculate the average channel value and then adjust each channel
# based on the average value and the difference from the theoretical average value of 128
# We multiply the average value by 100/255 to normalize the value to 0-100
# We then multiply the average value by 1.1 to increase the color balance
# Finally we convert the image back to BGR from LAB
# The above method works well for most images but can sometimes overcorrect
# For this reason we have also included a custom white balance option
# The custom white balance option allows the user to select the factor for each channel
# The custom white balance option is also useful for images that are already well balanced
# Resouces:
# https://stackoverflow.com/questions/46390779/automatic-white-balancing-with-grayworld-assumption
# https://stackoverflow.com/questions/24341114/simple-illumination-correction-in-images-opencv-c
# https://stackoverflow.com/questions/36982736/opencv-white-balancing-with-grayworld-assumption
# https://stackoverflow.com/questions/3490727/what-are-some-simple-methods-for-balancing-the-colors-of-an-image
if "AWB_custom" in options:
image = cv2.cvtColor(image, cv2.COLOR_BGR2LAB)
avg_a = np.average(image[:, :, 1])
avg_b = np.average(image[:, :, 2])
for x in range(image.shape[0]):
for y in range(image.shape[1]):
l, a, b = image[x, y, :]
# fix for CV correction
l *= 100 / 255.0
image[x, y, 1] = a - ((avg_a - 128) * (l / 100.0) * awb_b)
image[x, y, 2] = b - ((avg_b - 128) * (l / 100.0) * awb_g)
image = cv2.cvtColor(image, cv2.COLOR_LAB2BGR)
if "Blur" in options:
image = cv2.GaussianBlur(image, (0, 0), sigmablur)
if "Grayscale" in options:
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
image = cv2.cvtColor(image, cv2.COLOR_GRAY2BGR)
if "Denoise" in options:
image = cv2.bilateralFilter(image, diameter, sigmaColor, sigmaSpace)
if "Contrast" in options:
# Change contrast using a slider
contrast = st.slider('Contrast', 0.0, 10.0, 3.0)
image = cv2.convertScaleAbs(image, alpha=contrast, beta=0)
if "Brightness" in options:
image = cv2.convertScaleAbs(image, alpha=brightness, beta=0)
if "Gamma" in options:
# Perform gamma correction with a slider
image = np.power(image / 255.0, gamma)
image = cv2.convertScaleAbs(image, alpha=255.0)
if "Saturation" in options:
image = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
image[:, :, 1] = cv2.convertScaleAbs(image[:, :, 1], alpha=saturation, beta=0)
image = cv2.cvtColor(image, cv2.COLOR_HSV2BGR)
if "Sharpness" in options:
# sharpen the colors image
kernel = np.array([[-1,-1,-1], [-1,9,-1], [-1,-1,-1]])
image = cv2.filter2D(image, -1, kernel)
image = cv2.GaussianBlur(image, (0, 0), sigmaSharpness)
if "CLAHE" in options:
# convert from BGR to LAB color space
image = cv2.cvtColor(image, cv2.COLOR_BGR2LAB)
# split on 3 different channels
l, a, b = cv2.split(image)
# apply CLAHE to L-channel
clahe = cv2.createCLAHE(clipLimit=clipLimit, tileGridSize=(tileGridSize, tileGridSize))
l = clahe.apply(l)
# merge channels
image = cv2.merge((l, a, b))
# convert from LAB to BGR
image = cv2.cvtColor(image, cv2.COLOR_LAB2BGR)
if "Remove-Background" in options:
# Use the rembg library to remove the background
import rembg
# use a slider to adjust the alpha matting
image = rembg.remove(image, alpha_matting=alpha_matting, alpha_matting_foreground_threshold=fg_threshold, alpha_matting_background_threshold=bg_threshold)
if "Tile-Plakakia" in options:
from plakakia.utils_tiling import tile_image
tiles, coordinates = tile_image(image, tile_size, step_size)
# Draw a rectangle around each tile and make it smaller by 1 pixel to see the grid
tile_size = int(tile_size)
step_size = int(step_size)
for i in range(len(coordinates)):
x1, y1, x2, y2 = coordinates[i]
# Make the color the inverse of the average color of the tile
inv_color = (255 - int(np.average(tiles[i, :, :, 0])), 255 - int(np.average(tiles[i, :, :, 1])), 255 - int(np.average(tiles[i, :, :, 2])))
# Make the rectangle 1 pixel smaller to see the grid
cv2.rectangle(image, (x1, y1), (x2-1, y2-1), inv_color, 1)
st.write("Image size: ", image.shape)
if "Sobel_Vertical" in options:
if len(image.shape) == 3:
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
image = filters.sobel_v(image)
image = image / np.max(image)
image = (image * 255).astype(np.uint8)
if "Sobel_Horizontal" in options:
if len(image.shape) == 3:
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
image = filters.sobel_h(image)
image = image / np.max(image)
image = (image * 255).astype(np.uint8)
if image is not None:
cv2.imwrite("altered.png", image)
else:
st.error("Please initialize")
if image is not None:
if before_after:
from streamlit_image_comparison import image_comparison
st.header("Before and after")
image_comparison(img1="image.png", img2="altered.png")
else:
st.header("Processed image")
st.image(image, caption=f"", use_column_width=True, channels="BGR")
else:
st.stop()