StephaneBah's picture
v2
b6ec593 verified
import os
import gradio as gr
from PIL import Image
import numpy as np
import cv2
import matplotlib.pyplot as plt
import io
# Fonctions de traitement d'image
def load_image(image):
return image
def apply_negative(image):
img_np = np.array(image)
negative = 255 - img_np
return Image.fromarray(negative)
def grayscale(image):
return image.convert('L')
def binarize_image(image, threshold):
img_np = np.array(image.convert('L'))
_, binary = cv2.threshold(img_np, threshold, 255, cv2.THRESH_BINARY)
return Image.fromarray(binary)
def resize_image(image, width, height):
width = int(width)
height = int(height)
return image.resize((width, height))
def rotate_image(image, angle):
return image.rotate(angle)
def show_histogram(image):
image_gray = image.convert("L")
# Obtenir les données de l'image en niveaux de gris
image_array = np.array(image_gray)
# Calculer l'histogramme
hist, bins = np.histogram(image_array.flatten(), bins=256, range=[0,256])
# Créer une figure pour l'affichage de l'histogramme
fig, ax = plt.subplots()
ax.plot(hist, color='blue')
ax.set_xlim([0, 256])
ax.set_title('Histogram of Image')
# Enregistrer l'histogramme dans un buffer
buf = io.BytesIO()
plt.savefig(buf, format='png')
buf.seek(0)
# Ouvrir l'image du buffer en utilisant PIL
hist_image = Image.open(buf)
return hist_image
def gaussian_filter(image, shape=(3, 3)):
image = np.array(image)
filtered = cv2.GaussianBlur(image, shape, 0)
return Image.fromarray(filtered)
def mean_filter(image, shape=(3, 3)):
image = np.array(image)
filtered = cv2.blur(image, shape)
return Image.fromarray(filtered)
def sobel_edges(image, k=5):
k = int(k)
image = np.array(image.convert('L'))
sobel_x = cv2.Sobel(image, cv2.CV_64F, 1, 0, ksize=k)
sobel_y = cv2.Sobel(image, cv2.CV_64F, 0, 1, ksize=k)
sobel_combined = cv2.magnitude(sobel_x, sobel_y)
return Image.fromarray(np.uint8(sobel_combined))
def erosion(image, iterations=3, shape=(5, 5)):
iterations = int(iterations)
image = np.array(image.convert("L"))
kernel = np.ones(shape, np.uint8)
eroded_image = cv2.erode(image, kernel, iterations=iterations)
return Image.fromarray(eroded_image)
def dilatation(image, iterations=3, shape=(5, 5)):
iterations = int(iterations)
image = np.array(image.convert("L"))
kernel = np.ones(shape, np.uint8)
dilated_image = cv2.dilate(image, kernel, iterations=iterations)
return Image.fromarray(dilated_image)
# Interface Gradio
def image_processing(image, operation, modified_image, threshold=128, width=100, height=100, angle=30, k=5, iterations=3):
current_image = modified_image if modified_image is not None else image
if operation == "Négatif":
current_image = apply_negative(image)
elif operation == "Image en Gris":
current_image = grayscale(image)
elif operation == "Binarisation":
current_image = binarize_image(image, threshold)
elif operation == "Redimensionner":
current_image = resize_image(image, width, height)
elif operation == "Rotation":
current_image = rotate_image(image, angle)
elif operation == 'Filtre Gaussien':
current_image = gaussian_filter(image)
elif operation == 'Filtre Moyen':
current_image = mean_filter(image)
elif operation == 'Sobel Edges Extraction':
current_image = sobel_edges(image, k)
elif operation == 'Erosion':
current_image = erosion(image, iterations)
elif operation == 'Dilatation':
current_image = dilatation(image, iterations)
return current_image, show_histogram(current_image)
# Interface Gradio
with gr.Blocks() as demo:
gr.Markdown("## Traitement d'Images")
with gr.Row():
operation = gr.Radio(["Négatif", "Image en Gris", "Binarisation", "Redimensionner", "Rotation", 'Filtre Gaussien',
'Filtre Moyen', 'Sobel Edges Extraction', 'Erosion', 'Dilatation'], label="Opération", value="Négatif")
with gr.Row():
threshold = gr.Slider(0, 255, 128, label="Seuil de binarisation", visible=False)
width = gr.Number(value=100, label="Largeur", visible=False)
height = gr.Number(value=100, label="Hauteur", visible=False)
angle = gr.Slider(0, 360, 30, label="Angle de Rotation", visible=False)
k = gr.Number(value=5, label="k de Sobel", visible=False)
iterations = gr.Number(value=3, label="Nombre d'iteration pour les transformations morphologiques", visible=False)
def update_ui(operation):
# Mise à jour dynamique de la visibilité des champs
return {
threshold: gr.update(visible=operation == "Binarisation"),
width: gr.update(visible=operation == "Redimensionner"),
height: gr.update(visible=operation == "Redimensionner"),
angle: gr.update(visible=operation == "Rotation"),
k: gr.update(visible=operation == "Sobel Edges Extraction"),
iterations: gr.update(visible=operation in ["Erosion", "Dilatation"])
}
operation.change(update_ui, operation, [threshold, width, height, angle, k, iterations])
with gr.Row():
image_input = gr.Image(type="pil", label="Charger Image", scale=2)
original_hist = gr.Image(label="Histogramme de l'Image Originale", scale=1)
with gr.Row():
image_output = gr.Image(type="pil", label="Image Modifiée", interactive=False)
modified_hist = gr.Image(label="Histogramme de l'Image Modifiée", scale=1)
# Afficher l'histogramme de l'image d'entrée
def s_hist(image):
return show_histogram(image)
image_input.change(s_hist, inputs=image_input, outputs=original_hist)
submit_button = gr.Button("Appliquer")
submit_button.click(image_processing, inputs=[image_input, operation, image_output, threshold, width, height, angle, k, iterations], outputs=[image_output, modified_hist])
# Lancer l'application Gradio
demo.launch()