Enhancement of vascular visualization in laser speckle contrast imaging based on image algorithms

In practical biomedical applications, obtaining clear and focused speckle images through laser speckle contrast imaging (LSCI) presents significant challenges. These challenges are often compounded by motion artifacts and image noise, which can adversely affect the effectiveness of vascular visualization in LSCI.

We improved the visualization of blood flow in LSCI by focusing on three aspects: image registration, image denoising, and multi-focus image fusion.

We employed the Lucas-Kanade (LK) optical flow pyramid method alongside block matching and three-dimensional filtering (BM3D) algorithm based on guided filtering with total variation regularization to effectively mitigate motion artifacts and noise. Furthermore, we proposed a multi-focus image fusion technique based on the multi-scale image contrast amplification (MUSICA) algorithm, aimed at enhancing high-frequency signals and minimizing the effects of defocusing in LSCI.

The LK optical flow registration algorithm demonstrates improvements in both average peak signal-to-noise ratio and imaging quality compared with non-registration methods. The improved BM3D method outperforms classical denoising algorithms in various image evaluation parameters within LSCI. In the case of using the multi-focus image fusion method based on the MUSICA method, the image quality assessment of the sum of modulus of gray difference squared showed an improvement of nearly six times compared with the defocused images without the use of the MUSICA method.

Improvements in image processing algorithms, specifically in the areas of registration, denoising, and multi-focus image fusion, have significantly enhanced the visualization of blood flow in the vessels during practical applications of LSCI.