Nicole-Yi's picture
Upload 103 files
4a5e815 verified
import argparse
import cv2
import json
import os
from datetime import datetime
from imwatermark import WatermarkDecoder
from skimage.metrics import peak_signal_noise_ratio as compare_psnr
def evaluate_watermark(original_path, watermark_text, watermarked_path):
process_status = True
final_result_status = False
comments = []
# Timestamp
time_point = datetime.now().isoformat()
# Check input files
if not os.path.exists(original_path) or os.path.getsize(original_path) == 0:
comments.append(f"Error: Original image file '{original_path}' does not exist or is empty.")
process_status = False
if not os.path.exists(watermarked_path) or os.path.getsize(watermarked_path) == 0:
comments.append(f"Error: Watermarked image file '{watermarked_path}' does not exist or is empty.")
process_status = False
if process_status:
bgr_original = cv2.imread(original_path)
bgr_encoded = cv2.imread(watermarked_path)
if bgr_original is None or bgr_encoded is None:
comments.append("Error: Failed to read images, please check if files are corrupted or in correct format.")
process_status = False
if process_status:
try:
# decoder = WatermarkDecoder('bytes', len(watermark_text)*8)
# decoded_bytes = decoder.decode(bgr_encoded, 'dwtDct')
# extracted_text= decoded_bytes.decode('utf-8', errors='ignore')
# is_match = (extracted_text == watermark_text)
max_bits = 256
decoder = WatermarkDecoder('bytes', max_bits)
decoded_bytes = decoder.decode(bgr_encoded, 'dwtDct')
extracted_text = decoded_bytes.decode('utf-8', errors='ignore')
is_match = (watermark_text in extracted_text)
comments.append(f"{'✅' if is_match else '❌'} Extraction result: '{extracted_text}' | GT: '{watermark_text}'")
psnr_value = compare_psnr(bgr_original, bgr_encoded)
comments.append(f"📐 PSNR: {psnr_value:.2f} dB")
# Metrics
match_rate = '100%' if is_match else '0%'
psnr_satisfied = psnr_value >= 30.0
comments.append(f"🎯 Watermark detection_match: {match_rate}")
comments.append(f"🎯 PSNR ≥ 30.0: {'✅ Satisfied' if psnr_satisfied else '❌ Not satisfied'}")
final_result_status = is_match and psnr_satisfied
comments.append(f"Final evaluation result: Watermark match={is_match}, PSNR satisfied={psnr_satisfied}")
except Exception as e:
comments.append(f"Exception occurred during watermark processing or evaluation: {e}")
final_result_status = False
output_data = {
"Process": process_status,
"Result": final_result_status,
"TimePoint": time_point,
"Comments": "\n".join(comments)
}
print(output_data["Comments"])
return output_data
def write_to_jsonl(file_path, data):
"""
Append single result to JSONL file:
Each run appends one JSON line.
"""
try:
os.makedirs(os.path.dirname(file_path), exist_ok=True)
with open(file_path, 'a', encoding='utf-8') as f:
# Add default=str to handle non-serializable types with str()
f.write(json.dumps(data, ensure_ascii=False, default=str) + '\n')
print(f"✅ Result appended to JSONL file: {file_path}")
except Exception as e:
print(f"❌ Error occurred while writing to JSONL file: {e}")
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description="Extract and verify blind watermark, calculate image quality, and store results as JSONL")
parser.add_argument("--groundtruth", required=True, help="Path to original image")
parser.add_argument("--output", required=True, help="Path to watermarked image")
parser.add_argument("--watermark", required=True, help="Expected watermark content to extract")
parser.add_argument("--result", help="File path to store JSONL results")
args = parser.parse_args()
evaluation_result = evaluate_watermark(
args.groundtruth, args.watermark, args.output)
if args.result:
write_to_jsonl(args.result, evaluation_result)