Crystalline oxide semiconductors are promising back-end-of-line (BEOL)-compatible channel materials for AI hardware, yet their nanoscale trap physics remains unclear. Here, we directly mapped and quantified mobility (μ), trap density (<i>N</i><sub>eff</sub>), and trap depth in amorphous/nanocrystalline (a/n-) and polycrystalline (p-) In<sub>2</sub>O<sub>3</sub> films using scanning noise microscopy with finite-element analysis. A/n-In<sub>2</sub>O<sub>3</sub> exhibited large local variations in μ and <i>N</i><sub>eff</sub> with deep trap states (∼0.24 eV). Upon full crystallization, p-In<sub>2</sub>O<sub>3</sub> exhibited uniform μ and <i>N</i><sub>eff</sub> with shallow trap states at grains (∼0.10 eV) and grain boundaries (∼0.12 eV). Crystallization effectively eliminated structural-disorder-induced deep states, leaving only shallow donor-like oxygen vacancy traps. This led to enhanced μ and significantly reduced <i>N</i><sub>eff</sub> (and trap depth), exhibiting uniform spatial distributions with minute changes at grain boundaries. Furthermore, p-In<sub>2</sub>O<sub>3</sub> devices achieved higher mobility, more positive threshold voltage, and improved bias stability, confirming reduced deep-trap activity and enhanced charge-transport uniformity. This work establishes a direct link between structural ordering, local trap-depth modulation, and macroscopic electrical performances of crystalline oxide channels.