Robust Estimation of Upscaling Factor on Double JPEG Compressed Images

Area of Science:

• Digital image forensics
• Signal processing
• Information security

Background:

• Resampling detection is crucial in image forensics but faces challenges with JPEG compression robustness.
• JPEG compression degrades image content and introduces artifacts, complicating classical spectrum-based methods.

Purpose of the Study:

• To propose a novel method for estimating upscaling factors in double JPEG compressed images.
• To address the limitations of existing methods in handling JPEG compression and image upscaling.

Main Methods:

• Analysis of image spectrum to formulate the influence of scaling factors and JPEG compression parameters on artifacts.
• Derivation of expected peak positions in the frequency domain.
• Block Discrete Cosine Transform (BDCT) domain analysis for inverse scaling strategy.
• A fusion method combining frequency-domain and BDCT domain analysis for optimal upscaling factor estimation.

Main Results:

• The proposed method accurately estimates upscaling factors even with double JPEG compression and intermediate upscaling.
• Analytical derivation of characteristic peak positions provides a theoretical basis for detection.
• The fusion strategy effectively combines different domain analyses for improved accuracy.
• Experimental results show superior performance compared to existing state-of-the-art methods.

Conclusions:

• The developed method offers enhanced robustness for resampling detection in JPEG compressed images.
• This advancement contributes significantly to the field of image forensics, particularly for manipulated digital media.
• The fusion approach provides a more reliable estimation of upscaling factors in challenging scenarios.