Wavelet denoising of multiframe optical coherence tomography data.

Area of Science:

• Biomedical Imaging
• Signal Processing
• Optical Coherence Tomography (OCT)

Background:

• Speckle noise is a significant artifact in OCT images, degrading image quality and hindering analysis.
• Current speckle reduction methods often involve averaging multiple frames or denoising the final image, which can reduce resolution or require extensive data.

Purpose of the Study:

• To develop and evaluate a novel speckle noise reduction algorithm for OCT images.
• To improve signal-to-noise ratio (SNR) while preserving image sharpness and reducing the number of required input frames.

Main Methods:

• The algorithm utilizes wavelet decomposition of individual OCT frames for local noise and structure estimation.
• Wavelet detail coefficients are weighted, averaged, and reconstructed to achieve noise reduction.
• The method processes single frames, avoiding the need for simple averaging of multiple frames.

Main Results:

• Achieved approximately 100% signal-to-noise gain with only a minor 10.5% decrease in image sharpness (measured by full-width-half-maximum).
• The novel algorithm requires only 8 input frames to achieve comparable SNR gain to averaging 29 frames.
• Demonstrated effective differentiation between true tissue structures and speckle noise.

Conclusions:

• The proposed wavelet-based algorithm offers an efficient and effective method for speckle noise reduction in OCT images.
• This technique can serve as a valuable preprocessing step for retinal structure segmentation algorithms.
• The algorithm enhances OCT image quality with fewer frames and minimal sharpness loss, advancing biomedical imaging analysis.