Ultrasonic guided wave (UGW) imaging technology has been extensively adopted to assess the structural performance of composite laminates for its high detection accuracy, quick speed, and real-time detection capability. However, the algorithm design in composite structures is hindered by the complexity and anisotropy of dispersion. It is still challenging to optimize localization and detection range in sparse sensing arrays though the probabilistic diagnostic imaging (PDI) method avoids precise signal analysis. In this study, a modified delay-factor-based multiply–sum PDI (MDF-MSPDI) algorithm is examined. A novel ring-shaped probability distribution function based on the delay characteristics of damage scattering is defined, such that the off-axis localization and tolerance capabilities of the classical PDI method are enhanced. Subsequently, the delayed item is modified using the directional correlation of the group velocity in composite structures. Furthermore, given the spatial coherence of the path probability data, the new probability distribution terms are built through the nonlinear multiply–sum operation of path imaging to eliminate pointless damage artifacts and effects exerted by computational error. Finally, the algorithm performance was verified through simulated damage and impact damage experiments on composite laminates. As indicated by the result, the proposed algorithm exhibits excellent localization capability for damage in the internal and edge of the sensing arrays.