A table compares original, manipulated, and reconstructed images across four datasets.The table has five columns labeled: Dataset, Original Image, Manipulated Image, Reconstructed Image with modified weights, and Reconstructed Image with original weights. Row 1: Dataset: M I C C-F 2 2 0: The original image shows a snowy street scene. The manipulated image highlights a modified region with a red bounding box. The reconstructed image with modified weights appears slightly blurry and includes the metrics: P S N R equals 26.60 and S S I M equals 0.8973. The reconstructed image with original weights appears clearer and includes: P S N R equals 32.60 and S S I M equals 0.9347. Row 2: Dataset: Realistic-Tampering: The original image shows a pathway with trees. The manipulated image has a tampered vehicle marked with a red box. The reconstructed image with modified weights includes: P S N R equals 20.98 and S S I M equals 0.6492. The reconstructed image with original weights includes: P S N R equals 23.98 and S S I M equals 0.6970. Row 3: Dataset: Coverage row: The original image shows milk cartons on a supermarket shelf. The manipulated image shows an altered area highlighted with a red box. The reconstructed image with modified weights includes: P S N R equals 23.05 and S S I M equals 0.7988. The reconstructed image with original weights includes: P S N R equals 24.56 and S S I M equals 0.8288. Row 4: Dataset: High-Resolution row: The original image shows a car steering wheel with a small dog visible through the center. The manipulated image changes part of the steering area, marked with a red box. The reconstructed image with modified weights includes: P S N R equals 23.70 and S S I M equals 0.8405. The reconstructed image with original weights includes: P S N R equals 23.77 and S S I M equals 0.8413. |
