#!/usr/bin/env python3
"""عدسة النقوش — SAM + كلاسيكي"""
import cv2
import numpy as np
import base64, json, pickle, os
from http.server import HTTPServer, BaseHTTPRequestHandler
HTML_PAGE = """
عدسة النقوش - تفاعلي
⏳ يرجى الانتظار...
21
10
1
⚙️ جاري المعالجة...
"""
def process_image_bytes(img_bytes, min_val, max_val, thickness):
img_array = np.frombuffer(img_bytes, np.uint8)
img = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
if img is None: return None, "خطأ في قراءة الصورة"
h, w = img.shape[:2]
if w > 1000:
ratio = 1000 / w; img = cv2.resize(img, (1000, int(h * ratio)))
_, orig_buf = cv2.imencode('.jpg', img, [cv2.IMWRITE_JPEG_QUALITY, 85])
orig_b64 = base64.b64encode(orig_buf).decode()
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8, 8))
enhanced = clahe.apply(gray)
blurred = cv2.GaussianBlur(enhanced, (5, 5), 0)
bs = min_val if min_val % 2 == 1 else min_val + 1
thresh = cv2.adaptiveThreshold(blurred, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, bs, max_val)
kernel = np.ones((2, 2), np.uint8)
cleaned = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel, iterations=1)
white = cv2.bitwise_not(cleaned)
_, result_buf = cv2.imencode('.png', white)
result_b64 = base64.b64encode(result_buf).decode()
return {'original': orig_b64, 'result_img': result_b64}, None
def process_image_bytes_sam_init(img_bytes):
from segment_anything import sam_model_registry, SamPredictor
img_array = np.frombuffer(img_bytes, np.uint8)
img = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
if img is None: return None, "خطأ في قراءة الصورة"
h, w = img.shape[:2]
if w > 800:
ratio = 800 / w; img = cv2.resize(img, (800, int(h * ratio)))
_, orig_buf = cv2.imencode('.jpg', img, [cv2.IMWRITE_JPEG_QUALITY, 85])
orig_b64 = base64.b64encode(orig_buf).decode()
sam = sam_model_registry['vit_b'](checkpoint='/var/www/html/sam_vit_b_01ec64.pth')
predictor = SamPredictor(sam)
predictor.set_image(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
with open('/tmp/sam_predictor.pkl', 'wb') as f: pickle.dump(predictor, f)
with open('/tmp/sam_dims.pkl', 'wb') as f: pickle.dump({'img_h': img.shape[0], 'img_w': img.shape[1]}, f)
white = np.ones_like(img) * 255
cv2.imwrite('/tmp/sam_white.png', white)
_, white_buf = cv2.imencode('.png', white)
white_b64 = base64.b64encode(white_buf).decode()
# Reset history
for f in os.listdir('/tmp'):
if f.startswith('sam_hist_'): os.remove(f'/tmp/{f}')
cv2.imwrite('/tmp/sam_hist_0.png', white)
with open('/tmp/sam_count.txt', 'w') as f: f.write('0')
return {'original': orig_b64, 'white': white_b64}, None
def process_image_bytes_sam_click(img_bytes, click_x, click_y):
from segment_anything import SamPredictor
img_array = np.frombuffer(img_bytes, np.uint8)
img = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
if img is None: return None, "خطأ في قراءة الصورة"
h, w = img.shape[:2]
if w > 800:
ratio = 800 / w; img = cv2.resize(img, (800, int(h * ratio)))
with open('/tmp/sam_dims.pkl', 'rb') as f: dims = pickle.load(f)
px = int((click_x / 100.0) * dims['img_w'])
py = int((click_y / 100.0) * dims['img_h'])
with open('/tmp/sam_predictor.pkl', 'rb') as f: predictor = pickle.load(f)
masks, scores, logits = predictor.predict(point_coords=np.array([[px, py]]), point_labels=np.array([1]), multimask_output=False)
mask = masks[0]; mask_uint8 = (mask * 255).astype(np.uint8)
contours, _ = cv2.findContours(mask_uint8, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
white = cv2.imread('/tmp/sam_white.png')
if white is None: white = np.ones_like(img) * 255
cv2.drawContours(white, contours, -1, (0, 0, 0), 2)
cv2.imwrite('/tmp/sam_white.png', white)
# Save history
with open('/tmp/sam_count.txt', 'r') as f: cnt = int(f.read().strip())
cnt += 1
cv2.imwrite(f'/tmp/sam_hist_{cnt}.png', white)
with open('/tmp/sam_count.txt', 'w') as f: f.write(str(cnt))
_, result_buf = cv2.imencode('.png', white)
result_b64 = base64.b64encode(result_buf).decode()
return {'result_img': result_b64}, None
def process_image_bytes_sam_undo(step):
"""Load history at given step"""
hist_path = f'/tmp/sam_hist_{step}.png'
if not os.path.exists(hist_path): return None, "لا يوجد خطوات للتراجع"
white = cv2.imread(hist_path)
cv2.imwrite('/tmp/sam_white.png', white)
with open('/tmp/sam_count.txt', 'w') as f: f.write(str(step))
_, result_buf = cv2.imencode('.png', white)
result_b64 = base64.b64encode(result_buf).decode()
return {'result_img': result_b64}, None
def process_image_bytes_sam_erase(img_bytes, click_x, click_y):
"""Erase at click point: draw white over the mask area"""
from segment_anything import SamPredictor
img_array = np.frombuffer(img_bytes, np.uint8)
img = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
if img is None: return None, "خطأ في قراءة الصورة"
h, w = img.shape[:2]
if w > 800:
ratio = 800 / w; img = cv2.resize(img, (800, int(h * ratio)))
with open('/tmp/sam_dims.pkl', 'rb') as f: dims = pickle.load(f)
px = int((click_x / 100.0) * dims['img_w'])
py = int((click_y / 100.0) * dims['img_h'])
with open('/tmp/sam_predictor.pkl', 'rb') as f: predictor = pickle.load(f)
masks, scores, logits = predictor.predict(point_coords=np.array([[px, py]]), point_labels=np.array([1]), multimask_output=False)
mask = masks[0]; mask_uint8 = (mask * 255).astype(np.uint8)
contours, _ = cv2.findContours(mask_uint8, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
white = cv2.imread('/tmp/sam_white.png')
if white is None: white = np.ones_like(img) * 255
# Draw WHITE over the area to erase (opposite of click)
cv2.drawContours(white, contours, -1, (255, 255, 255), -1) # filled white
cv2.imwrite('/tmp/sam_white.png', white)
with open('/tmp/sam_count.txt', 'r') as f: cnt = int(f.read().strip())
cnt += 1
cv2.imwrite(f'/tmp/sam_hist_{cnt}.png', white)
with open('/tmp/sam_count.txt', 'w') as f: f.write(str(cnt))
_, result_buf = cv2.imencode('.png', white)
result_b64 = base64.b64encode(result_buf).decode()
return {'result_img': result_b64}, None
class Handler(BaseHTTPRequestHandler):
def do_GET(self):
self.send_response(200)
self.send_header('Content-Type', 'text/html; charset=utf-8')
self.end_headers()
self.wfile.write(HTML_PAGE.encode())
def do_POST(self):
ct = self.headers['Content-Type']; cl = int(self.headers['Content-Length'])
body = self.rfile.read(cl); boundary = ct.split('=')[1].encode()
parts = body.split(b'--' + boundary)
path = self.path
if path == '/process':
img_bytes = None; min_val = 21; max_val = 10; thickness = 1
for p in parts:
if b'Content-Disposition' not in p: continue
if b'name="image"' in p: i = p.find(b'\r\n\r\n'); img_bytes = p[i+4:].rstrip(b'\r\n--') if i!=-1 else None
elif b'name="min_val"' in p: i = p.find(b'\r\n\r\n'); min_val = int(p[i+4:].rstrip(b'\r\n--').decode()) if i!=-1 else min_val
elif b'name="max_val"' in p: i = p.find(b'\r\n\r\n'); max_val = int(p[i+4:].rstrip(b'\r\n--').decode()) if i!=-1 else max_val
elif b'name="thickness"' in p: i = p.find(b'\r\n\r\n'); thickness = int(p[i+4:].rstrip(b'\r\n--').decode()) if i!=-1 else thickness
if not img_bytes: self._json({'error': 'لم يتم العثور على الصورة'}); return
r, e = process_image_bytes(img_bytes, min_val, max_val, thickness)
if e: self._json({'error': e}); return
self._json(r)
elif path == '/sam_init':
img_bytes = None
for p in parts:
if b'Content-Disposition' in p and b'name="image"' in p:
i = p.find(b'\r\n\r\n')
if i != -1: img_bytes = p[i+4:].rstrip(b'\r\n--')
break
if not img_bytes: self._json({'error': 'لم يتم العثور على الصورة'}); return
r, e = process_image_bytes_sam_init(img_bytes)
if e: self._json({'error': e}); return
self._json(r)
elif path == '/sam_click':
img_bytes = None; cx = 50.0; cy = 50.0
for p in parts:
if b'Content-Disposition' not in p: continue
if b'name="image"' in p: i = p.find(b'\r\n\r\n'); img_bytes = p[i+4:].rstrip(b'\r\n--') if i!=-1 else None
elif b'name="x"' in p: i = p.find(b'\r\n\r\n'); cx = float(p[i+4:].rstrip(b'\r\n--').decode()) if i!=-1 else cx
elif b'name="y"' in p: i = p.find(b'\r\n\r\n'); cy = float(p[i+4:].rstrip(b'\r\n--').decode()) if i!=-1 else cy
if not img_bytes: self._json({'error': 'لم يتم العثور على الصورة'}); return
r, e = process_image_bytes_sam_click(img_bytes, cx, cy)
if e: self._json({'error': e}); return
self._json(r)
elif path == '/sam_undo':
step = 0
for p in parts:
if b'Content-Disposition' in p and b'name="step"' in p:
i = p.find(b'\r\n\r\n')
if i != -1: step = int(p[i+4:].rstrip(b'\r\n--').decode())
break
r, e = process_image_bytes_sam_undo(step)
if e: self._json({'error': e}); return
self._json(r)
elif path == '/sam_erase':
img_bytes = None; cx = 50.0; cy = 50.0
for p in parts:
if b'Content-Disposition' not in p: continue
if b'name="image"' in p: i = p.find(b'\r\n\r\n'); img_bytes = p[i+4:].rstrip(b'\r\n--') if i!=-1 else None
elif b'name="x"' in p: i = p.find(b'\r\n\r\n'); cx = float(p[i+4:].rstrip(b'\r\n--').decode()) if i!=-1 else cx
elif b'name="y"' in p: i = p.find(b'\r\n\r\n'); cy = float(p[i+4:].rstrip(b'\r\n--').decode()) if i!=-1 else cy
if not img_bytes: self._json({'error': 'لم يتم العثور على الصورة'}); return
r, e = process_image_bytes_sam_erase(img_bytes, cx, cy)
if e: self._json({'error': e}); return
self._json(r)
def _json(self, data):
self.send_response(200)
self.send_header('Content-Type', 'application/json')
self.end_headers()
self.wfile.write(json.dumps(data).encode())
def log_message(self, *a): pass
if __name__ == '__main__':
port = 8085
print(f"✍️ عدسة النقوش على http://localhost:{port}")
HTTPServer(('0.0.0.0', port), Handler).serve_forever()