Noise reduction by adaptive-SIN filtering for retinal OCT images

Area of Science:

• Ophthalmology
• Medical Imaging
• Image Processing

Background:

• Optical Coherence Tomography (OCT) is vital for ophthalmic examination.
• Image quality in OCT is often degraded by noise, impacting diagnostic accuracy.
• Existing denoising methods struggle with the specific noise characteristics of OCT.

Purpose of the Study:

• To develop an adaptive denoising algorithm for OCT images.
• To address the limitations of standard denoising techniques when applied to OCT data.
• To improve the preservation of image details during noise reduction.

Main Methods:

• A square-root transform was employed to redistribute OCT noise, approximating a Gaussian distribution.
• An adaptive 3D Shearlet filter with a noise-redistribution scheme (adaptive-SIN) was proposed.
• The algorithm was evaluated on three benchmark datasets using quantitative and qualitative assessments.

Main Results:

• The adaptive-SIN algorithm effectively transformed Poisson noise to Gaussian noise for optimal Shearlet transform application.
• The adaptive thresholding scheme successfully adapted to varying noise conditions.
• The proposed method demonstrated superior noise removal and detail preservation compared to eight other algorithms.

Conclusions:

• The adaptive-SIN algorithm offers a significant advancement in denoising OCT images.
• The method provides improved diagnostic quality by effectively removing noise while preserving fine image structures.
• This approach holds promise for enhancing the reliability and utility of OCT in clinical practice.