Surface thermal lensing (STL) is a high sensitive pump-probe technique for photothermal characterization of weakly absorbing defects in optical materials. Commonly, the frequency modulation is applied in the pump beam, and the STL signal that appears as small modulation on top of an intense background is then detected on the probe beam by lock-in amplification. However, the lock-in detection by filtering the signal in the frequency domain results in low efficiency in the STL, preventing investigations of the large area surfaces. Here it is shown that the multi-channel averaging, which corresponds to temporal integration, is better suited for detection of the STL signal. It is demonstrated that by converting the conventional single-point scanning STL to the line scanning STL and employing suitable width of the time bins and the number of records, the imaging speed of the multi-channel averaging detection can be improved by a factor of 3.7 relative to the lock-in detection at the same signal-to-noise ratio of 6 dB and imaging resolution of 20 μm. Therefore, high-quality photothermal images of weakly absorbing defects can be recorded with higher efficiency and lower laser irradiance, making it possible to investigate large samples.