Multiresolution image representation using combined 2-D and 1-D directional filter banks.

Area of Science:

• Image processing
• Signal processing
• Computer vision

Background:

• Existing multiresolution image representations often involve complex implementations or high computational costs.
• Conventional methods like separable wavelet transforms (WT) and contourlets have limitations in efficiency and performance.

Purpose of the Study:

• To present effective multiresolution image representations.
• To develop methods with simple implementation and low computational costs.
• To achieve nonredundant transforms for quad-tree like representations.

Main Methods:

• Combining 2-D filter banks (FB) with directional wavelet transform (WT).
• Developing nonredundant transforms for quad-tree like multiresolution representations.
• Applying proposed filter banks to nonlinear approximation, image coding, and denoising.

Main Results:

• The proposed methods offer simpler implementation and lower computation costs compared to existing techniques.
• Achieved nonredundant transforms enabling quad-tree like multiresolution representations.
• Demonstrated visual quality improvements and higher Peak Signal-to-Noise Ratio (PSNR) in applications.

Conclusions:

• The proposed filter bank and directional wavelet transform combination provides an effective and efficient approach for multiresolution image representation.
• These methods outperform conventional separable WT and contourlet in key image processing tasks.
• The techniques are suitable for applications requiring high visual quality and computational efficiency.