Lossless Recompression of JPEG Images Using Transform Domain Intra Prediction.

Area of Science:

• Computer Science
• Image Processing
• Data Compression

Background:

• JPEG is a widely used image format, but its compression efficiency lags behind modern methods.
• The proliferation of digital images increases storage costs, necessitating efficient compression techniques.
• Lossless recompression of existing JPEG images is crucial for reducing storage while preserving fidelity.

Purpose of the Study:

• To propose a novel hybrid coding framework, LLJPEG, for lossless recompression of JPEG images.
• To enhance storage efficiency for JPEG images through advanced compression techniques.
• To address the limitations of current JPEG compression performance.

Main Methods:

• A hybrid coding framework (LLJPEG) utilizing transform domain intra prediction.
• Implementation of block partition, intra prediction, Discrete Cosine Transform (DCT), quantization, and entropy coding.
• Redesign of entropy coding and introduction of new intra prediction tools.

Main Results:

• LLJPEG achieves significant storage reduction: 29.43% on Kodak and 26.40% on DIV2K datasets.
• The method ensures lossless compression, maintaining original image fidelity.
• Low decoding complexity was observed during experimental evaluations.

Conclusions:

• LLJPEG offers a highly effective solution for lossless recompression of JPEG images.
• The proposed framework significantly reduces storage requirements for JPEG data.
• LLJPEG presents a viable approach for efficient image storage and transmission.