Denoising algorithm of OCT images via sparse representation based on noise estimation and global dictionary

Area of Science:

• Medical Imaging
• Optical Engineering
• Signal Processing

Background:

• Optical Coherence Tomography (OCT) is a vital non-invasive imaging technique.
• Speckle noise significantly degrades OCT image quality due to light's low-coherent interference.
• Existing denoising methods often struggle with noise reduction and edge preservation.

Purpose of the Study:

• To develop an efficient OCT image denoising algorithm.
• To improve OCT image quality by reducing speckle noise.
• To enhance edge preservation and computational efficiency in OCT image processing.

Main Methods:

• A novel denoising algorithm based on sparse representation is proposed.
• A global dictionary is constructed from high-quality OCT training images.
• Noise intensity is estimated for each input image, followed by sparse decomposition and reconstruction.

Main Results:

• The proposed algorithm effectively removes speckle noise from OCT images.
• High-quality OCT images with preserved edges are achieved.
• The method demonstrates superior performance compared to adaptive dictionary and other denoising algorithms, with reduced computational cost.

Conclusions:

• The developed sparse representation-based algorithm offers an efficient solution for OCT image denoising.
• It significantly improves image quality, particularly for images with varying noise intensities.
• The algorithm provides a valuable tool for enhancing OCT imaging applications.