Corrosion and cracks are common defects in structural plates. The mode conversion of Lamb waves at these non-perforating damages is a primary factor limiting the quality of Lamb wave imaging. Meanwhile, acoustic diffraction adheres to the Rayleigh criterion, leading to resolution limits in ultrasonic imaging. This paper designed a fully convolutional network to segment and reconstruct the received signals, enabling the automatic extraction of target modes and eliminating interference from clutter and mode conversions. Additionally, a sign coherence factor-total focusing method ( SCF-TFM) is proposed, where the symbolic coherence factor is applied during the total focus method imaging process, suppressing the interference from scattered waves in non-target regions. By considering both amplitude and phase information of the signals, it can partially overcome the limitations of the Rayleigh criterion, achieving super-resolution imaging. Experimental results demonstrate that for a single blind-hole defect, the lateral resolution of the imaging result using this method is 62. 41% higher than that of total focus method, and the signal-to-noise ratio ( SNR) is increased by 58. 23% . For multiple asymmetric blind-hole defects, when the spacing between defects exceeds the Rayleigh resolution limit, the signal-to-noise ratio improves by 92. 89% using this method. When the spacing is below the Rayleigh resolution limit, this method can achieve super-resolution imaging. Keywords:Lamb waves; asymmetric blind hole defects; ful