Abstract: The reduction in seismic resolution caused by earth absorption, known as Q effect, is characterized by frequency-dependent amplitude attenuation and velocity dispersion. Inverse Q-filtering serves as a fundamental technique for attenuation compensation to enhance seismic resolution. However, high-frequency noises pose a trade-off for conventional methods as enhancing resolution inevitably degrades signal-to-noise ratio, thereby limiting their performance. To address this issue, we propose a multi-trace absorption compensation framework for the accurate stable recovery of seismic high-frequency components. A signal identification operator, expressed as a prediction error filter, is derived from attenuation data by leveraging the intrinsic mapping between low-frequency and high-frequency components. This operator captures inter-frequency dependencies and is integrated into the inversion objective function as a robust constraint for the precise recovery of high-frequency structures. Synthetic and field tests demonstrate the method's performance in achieving high-resolution seismic data with an improved signal-to-noise ratio.