Parallel Dual-Path Network for Clarity Restoration of Scanned Posters via Frequency Decomposition

Scanned poster materials frequently exhibit blur degradation arising from scanner hardware constraints, document manipulation movements, and optical apparatus imperfections. This research introduces PDPNet, a parallel dual-path network framework designed for clarity recovery in digitized graphic documents, with particular emphasis on poster and print media scenarios. The architecture operates via two autonomous processing channels: a spatial pathway preserving layout organization and textual content integrity, alongside a frequency pathway recovering fine printing details through wavelet-based decomposition mechanisms. An adaptive integration strategy dynamically merges outputs from both channels according to regional content characteristics. Experimental assessments on poster scanning collections demonstrate that our methodology attains PSNR of 31.82 dB (mean over 3 runs) and SSIM of 0.943, constituting statistically significant improvements (p