Abstract: In the monitoring, diagnosis, and clinical management of fundus diseases, the visualization of fundus vascular structures is recognized as of paramount importance for accurate pathological assessment. Laser speckle contrast imaging, a well-established non-invasive optical imaging technique, is capable of achieving vessel visualization and real-time dynamic blood flow monitoring, as contrast agents are not required. However, significant noise is inevitably introduced by the conventional contrast calculation method that employs a fixed-size sliding window. In the present study, laser speckle contrast images are aligned and fused with vascular structural images acquired via LED light illumination. Subsequent to fusion, the composite images are segmented into regional subsets, and region-specific optimal pixels are selected based on the variance criterion to perform contrast computation. Within the vascular boundary region, the geometry and dimensions of the analysis window are dynamically adjusted to adapt to the local vascular morphology. The results indicate that the contrast has been enhanced by 67.61%. Moreover, in deeper vascular imaging, the proposed method is found to yield superior noise reduction performance and enhanced image quality, thereby providing support for the diagnosis and management of ocular diseases.
Key words: laser speckle contrast imaging; fundus; blood flow; image registration; adaptive window
（Acta Laser Biology Sinica, 2025, 34(6): 541-550）