Limited-view photoacoustic (PA) imaging suffers from incomplete acoustic wave acquisition due to a limited number of sensor configurations. This results in degraded image quality and makes it difficult to accurately depict microvascular networks. The challenge is more pronounced in optical-resolution photoacoustic microscopy (OR-PAM), where weak vascular signals and noise often result in fragmented, discontinuous vessel representations. To address these limitations, we proposed a method that enhances weak vascular signals from raw limited-view OR-PAM images without any postprocessing steps, enabling effective noise suppression and fine vessel feature extraction. We present a segmentation framework that combines a Modified Weak Signal Attenuation (MWSA) module with a self-attention mechanism (SAM) optimized using Dice and Binary Cross-Entropy (BCE) loss functions with a Multi-scale Perception (MP) module. Validation on OR-PAM datasets of mouse ear vessel models shows that our method improves vessel continuity and segmentation accuracy over conventional techniques. The results obtained are superior to the previous method that combines WSA with a Spatial Attention Module. With higher accuracy, sensitivity, and Dice scores, our method offers a reliable solution for microvasculature extraction and has strong potential for vascular health studies and diagnostic applications.