JPEG compression history estimation for color images.

Area of Science:

• Digital Image Processing
• Computer Vision
• Signal Processing

Background:

• Digital color images are frequently encountered after Joint Photographic Experts Group (JPEG) compression.
• JPEG compression settings (compression history or CH) are often lost after decompression.
• Recovering lost CH is crucial for effective image recompression.

Purpose of the Study:

• To estimate the lost JPEG compression history (CH) of a JPEG-decompressed color image.
• To develop robust algorithms for JPEG Compression History Estimation (CHEst).
• To demonstrate the utility of estimated CH in optimizing JPEG recompression.

Main Methods:

• Observation of lattice structure in the Discrete Cosine Transform (DCT) domain introduced by JPEG quantization.
• Development of a statistical dictionary-based CHEst algorithm using maximum a posteriori estimation.
• Design of a blind lattice-based CHEst algorithm exploiting 3-D parallelepiped lattice structures in DCT coefficients.

Main Results:

• Both proposed CHEst algorithms demonstrate robust performance in practice.
• The statistical dictionary-based method tests various CHs against a dictionary.
• The blind lattice-based method utilizes novel lattice algorithms for CH estimation.
• Simulations confirm the effectiveness of JPEG CHEst in recompression scenarios.

Conclusions:

• Estimated JPEG CH enables recompression with high signal-to-noise ratio (minimal distortion).
• The recovered CH facilitates achieving smaller file sizes during recompression.
• The developed CHEst algorithms offer practical solutions for optimizing JPEG image recompression.