Subregion-based maximum similarity search registration method for transverse chromatic aberration correction in ultra-widefield scanning laser ophthalmoscopy images.

Abstract

Significance: Ultra-widefield scanning laser ophthalmoscopy (UWF SLO), a confocal scanning-based ophthalmic imaging modality widely used in clinical and research ophthalmology, offers an ultra-wide field of view, high resolution, and real-time dynamic imaging. However, multiwavelength imaging-induced transverse chromatic aberration (TCA) is markedly exacerbated in the peripheral retina, causing ghosting artifacts of retinal vessels and other critical anatomical structures. This severely limits image quality and the reliability of downstream quantitative analysis.

Aim: We aim to correct TCA in multiwavelength UWF SLO images using a subregion-based maximum similarity search registration method, thereby eliminating vascular ghosting artifacts to enhance image quality for reliable clinical analysis.

Approach: In this study, we address the nonuniform TCA between the red and green channels in multiwavelength UWF SLO imaging through three core steps: grid-based subregion sampling, high-precision control point extraction via local optimal matching guided by zero-mean normalized cross-correlation (ZNCC), and global perspective transformation model fitting. The effectiveness of the proposed method was validated on fundus image samples acquired by our in-house-developed UWF SLO system, through both qualitative visual assessment and quantitative evaluation metrics.

Results: The proposed correction method significantly mitigated TCA in UWF SLO images and greatly improved the spatial alignment of retinal vascular contours between the red and green channels. Validation on 11 fundus images showed a significant enhancement in vascular matching performance: the mean dice similarity coefficient increased by 26.7% relatively (from $0.595\pm 0.060$ to $0.754\pm 0.035$ ) and the mean intersection over union increased by 42.5% relatively (from $0.426\pm 0.061$ to $0.607\pm 0.044$ ). Paired samples t-test verified highly statistically significant improvements in both metrics (all $p<0.001$ ).

Conclusions: The proposed subregion-based maximum similarity search registration method effectively corrects TCA in multiwavelength UWF SLO fundus images, significantly improves image quality, and provides reliable technical support for clinical ophthalmic practice and downstream quantitative fundus image analysis.