Wavelet-based image coding using nonlinear interpolative vector quantization

Area of Science:

• Digital image processing
• Signal processing
• Computer vision

Background:

• Wavelet-based image coding is crucial for efficient data compression.
• Existing methods often face challenges with computational complexity.
• Vector quantization (VQ) is a powerful technique for data compression.

Purpose of the Study:

• To develop a reduced complexity wavelet-based image coding technique.
• To improve coding performance through efficient feature extraction and VQ.
• To enable efficient reconstruction of image data at the decoder.

Main Methods:

• Formation of 64-D vectors from wavelet subimage coefficients across three decomposition stages.
• Extraction of 16-D feature vectors from the 64-D vectors.
• Application of vector quantization (VQ) on the extracted feature vectors.
• Reconstruction of 64-D vectors at the decoder using nonlinear interpolation.

Main Results:

• The proposed technique significantly reduces computational complexity compared to standard methods.
• Potential for superior coding performance is demonstrated, especially when codebooks are trained on similar images.
• Efficient reconstruction of image data is achieved through nonlinear interpolation.

Conclusions:

• The developed wavelet-based image coding technique offers a practical solution for reduced complexity.
• The method shows promise for high-performance image compression applications.
• Codebook generation strategy is key to optimizing performance for specific image types.