Optical time-domain reflectometry (OTDR) is widely used for fault detection and optical path loss characterization in optical fiber systems and networks. While long pulses are necessary for adequate signal-to-noise ratio and measurement range, they blur localized fault events, degrading both spatial resolution and the ability to detect non-reflective or weakly reflective localized losses. For short-distance optical systems that cascade multiple optical components where spatial resolution in the 10 cm to 1 m range is desired, this limitation is particularly significant. This work addresses these constraints through developing blind pulse spread function (PSF) estimation and post-processing of OTDR traces. Our method exploits localized reflection events (such as induced reference reflections at known locations) to estimate the system's impulse response without requiring hardware modifications or independent pulse characterization. The estimated PSF is then used to perform deconvolution, along with noise reduction processing, on OTDR traces acquired from commercial instruments on laboratory systems. By iteratively refining PSF estimates and applying advanced signal processing techniques adapted from image deconvolution literature, we aim to enhance spatial resolution and improve detection fidelity for discrete loss events. We will present preliminary results from OTDR trace simulations demonstrating the feasibility of blind PSF estimation using iterative optimization methods. Ongoing measurements using commercial OTDR instruments on laboratory systems will be conducted to validate the approach on real-world data.