First-person videos (FPVs) captured by wearable cameras have undesired shakiness because of fast changing views. When existing video stabilization techniques are applied, FPVs are transformed into cinematographic videos, losing the First-person motion information (FPMI) such as the recorder's interests and actions. We propose a system that can enhance viewability of FPVs by stabilizing them while preserving their FPMI. The viewability is charaterized based on a human perception model. Objective tests show that our method has competitive stabilization performance relative to existing video stabilization techniques. And subjective tests show that spectators still experience the FPMI from the resulting videos while shakiness is reduced.