Adaptive 2-D wavelet transform based on the lifting scheme with preserved vanishing moments

Area of Science:

• Digital Image Processing
• Signal Processing
• Computer Vision

Background:

• Wavelet filter banks are crucial for image processing tasks like denoising.
• Existing methods often lack adaptability to local image features.
• Nonseparable filter banks offer advantages in certain applications.

Purpose of the Study:

• To propose novel adaptive wavelet filter bank structures.
• To enhance image denoising performance by adapting to local image features.
• To retain essential properties like vanishing moments in adaptive filters.

Main Methods:

• Development of nonseparable wavelet filter banks based on quincunx sampling.
• Pixel-wise adaptation of filter bank properties using the intersection of confidence intervals (ICI) rule.
• Integration of adaptation within the predict stage of the lifting scheme.

Main Results:

• The proposed adaptive filter banks successfully adapt to local image characteristics.
• The method retains desirable primal and dual vanishing moments.
• Demonstrated superior performance in denoising synthetic images with periodic patterns compared to state-of-the-art methods.

Conclusions:

• The novel adaptive wavelet filter banks offer significant improvements in image denoising.
• The pixel-wise adaptation strategy effectively handles local image features.
• The method shows particular promise for denoising images with periodic structures.