Statistical model for OCT image denoising

Area of Science:

• Biomedical Imaging
• Image Processing
• Medical Technology

Background:

• Optical coherence tomography (OCT) is vital for non-invasive imaging in clinical and biological applications.
• Speckle noise in OCT images degrades image quality, hindering accurate analysis and diagnosis.
• Existing denoising methods may compromise image details or computational efficiency.

Purpose of the Study:

• To introduce a novel algorithm for denoising spectral domain OCT (SD-OCT) images.
• To improve the diagnostic utility of OCT by effectively reducing speckle noise.
• To preserve image edges and reduce computational cost during the denoising process.

Main Methods:

• Developed a denoising algorithm based on maximum-a-posteriori (MAP) estimation.
• Incorporated a new speckle noise model derived from local statistics of SD-OCT data.
• Utilized a Huber variant of total variation regularization for enhanced edge preservation.

Main Results:

• The proposed algorithm successfully reduced speckle noise in OCT images.
• Edge preservation was effectively maintained, ensuring structural details were not lost.
• The method achieved these results at a reduced computational cost compared to existing approaches.

Conclusions:

• The novel OCT denoising algorithm offers a significant improvement for image analysis and diagnostic applications.
• The combination of a new noise model and advanced regularization effectively addresses speckle noise while preserving image fidelity.
• This efficient algorithm presents a valuable tool for enhancing OCT imaging in various scientific and medical fields.