Metallic engineering components exposed to high temperature for an extensive time are vulnerable to deterioration of mechanical properties due to the formation of micro-structural defects such as creep, embrittlement, graphitization. These defects starts to initiate when the temperature of engineering components is greater than a threshold. As a useful tool, thermography can assist in providing relevant information for the determination of locations where these defects can potentially form.Conventional thermography methods in non-destructive testing, for example, the infrared thermography, are applicable only to map the surface temperature which could not estimate the positions of the subsurface defects caused by high temperature. There has been some development of internal thermal imaging techniques in the medical field based on magnetic-resonance (MR) and X-ray. However, these methods are not applicable to high-temperature metallic materials in NDT considering their limited measuring range of less than 100 • C, the low penetration depth, and potential disturbance to the original temperature field.The work presented in this thesis extends the scope of temperature imaging for hightemperature metals from surface to subsurface. The challenge of the subsurface thermography comes from satisfying simultaneously the four requirements: 1) to avoid damages of measuring instruments caused by the high temperature of the heated sample, 2) compatible with metal materials, 3) little disturbance to the target temperature field, 4) reliable data collection with high SNR. To overcome the challenge, the non-contact subsurface thermography using laser-ultrasonic waveform ix ABSTRACT inversion is proposed.The subsurface thermography based on the traveltimes of Rayleigh wave is presented. The feasibility of the proposed thermography is validated experimentally.A relative error of 10% within 1 wavelength is observed. The performance of the proposed subsurface thermography, including the average error, the reconstruction depth and the resistance to noise, is investigated numerically. The effective range of temperature reconstruction is defined, within which the average reconstruction error is around 5%. The depth of effective reconstruction is around 1.3 wavelength.The inverted results with noise added at different levels indicate that the proposed method has a strong resistance to noise.Apart from traveltime, other information such as frequency and amplitude, contained in the laser-ultrasound waveforms can also provide useful information for subsurface thermography. The subsurface thermography based on the full-waveform inversion method is therefore proposed. The experimental results show a relative error of 10% within a depth of 1 wavelength. Compared with the experimental results using traveltime, the overall accuracy of the experimental results is lower as full-waveform inversion requires a higher accuracy in the collected signals and the consistency between the numerical configuration and the experimental setup. Numerical models ...