In the field of laser ultrasonic non-destructive testing for surface defects, this paper addresses the significant challenge posed by the high sensitivity of Lamb wave signals to noise interference, which severely compromises defect localization accuracy. To overcome this limitation, a novel surface defect localization method is proposed, integrating CPO-optimized variational mode decomposition (VMD) denoising with a dual-feature energy averaging strategy. The approach begins by employing the Crested Porcupine Optimizer (CPO) to adaptively optimize the mode number and penalty factor in VMD, thereby enabling optimal modal decomposition of complex Lamb wave signals. Subsequently, the Permutation Entropy (PE) criterion is utilized to assess the decomposed intrinsic mode functions (IMF), facilitating effective discrimination between noise-dominated and signal-dominant components, followed by signal reconstruction and noise suppression. In addition, a dual-feature averaging strategy using signal energy as the evaluation metric is introduced to improve both the accuracy and robustness of defect localization. Simulation results demonstrate that, under low signal-to-noise ratio (SNR) conditions and across multiple measurement points, the proposed CPO-VMD method outperforms conventional approaches including empirical mode decomposition (EMD), standard VMD, and complementary ensemble empirical mode decomposition (CEEMDAN) in preserving key time-domain characteristics of Lamb waves. Experimental validation is performed using laser ultrasonic data collected from three test cases with varying scanning distances and crack positions. Denoising is conducted using CEEMDAN, EMD, VMD, and CPO-VMD, respectively, after which defect localization is implemented using the dual-feature averaging method. The results show that the CPO-VMD-based framework achieves the lowest relative localization errors of 1.62%, 1.59%, and 2.35%, confirming its superior performance over comparative methods. The methodology presented in this study demonstrates considerable potential for practical deployment in laser ultrasonic non-destructive testing of surface defects.