Effective Decompression of JPEG Document Images.

Area of Science:

• Digital Image Processing
• Computer Vision
• Information Theory

Background:

• JPEG compression is widely used but can cause significant distortion at low bit rates, particularly for document images.
• Existing decompression methods may struggle with time-critical applications on mobile networks.
• Document images contain distinct regions (smooth backgrounds, sharp text/graphics) requiring specialized processing.

Purpose of the Study:

• To develop an effective and efficient JPEG decompression approach for document images.
• To optimize decompression for time-critical applications, especially on mobile infrastructures.
• To improve the visual quality of heavily distorted document images compressed at low bit rates.

Main Methods:

• Image blocks are classified into smooth and non-smooth categories.
• Smooth blocks are decoded in the transform domain by minimizing boundary variation.
• Non-smooth blocks are reconstructed using a novel text model and an efficient optimization algorithm.

Main Results:

• The proposed method significantly improves the visual quality of decompressed JPEG document images.
• The approach demonstrates effectiveness even when images are encoded at very low bit rates.
• Experimental validation confirms the approach's performance and efficiency.

Conclusions:

• The developed decompression technique offers a superior solution for handling distorted JPEG document images.
• This method is particularly beneficial for mobile applications requiring fast and high-quality image reconstruction.
• The block classification and specialized decoding strategies contribute to enhanced visual fidelity.