Abstract: In the two-dimensional results in ultrasound thermal strain imaging (TSI), intense noise may appear near (in the direction away from the probe) the heating region due to the thermo-acoustic lens effect and peak mislocalization in the cross-correlation function, making identification of the region of interest (ROI) a challenging task. An integrated TSI method based on frequency domain segmentation is proposed to achieve noise reduction in TSI images and optimize the selection of the ROIs. Firstly, ultrasonic RF data during the heating process is divided into multiple frequency bands, while data in each band is treated as the result of an independent imaging process. Then, TSI calculations are carried out in each band. The morphologies of ROIs and thermal spots are similar for different bands, while the noise distribution varies. Based on this observation, the ROI can be easily distinguished from the noise region, and the integration of multi-band TSI realizes noise reduction. Finally, the radial gradient index is used to identify the ROI. The proposed method is verified via numerical simulations and in vitro experiments. The contrast in the ROI of the accumulated thermal strain image is raised by 1.7 dB in the simulation and by 13.6 dB in the experiment.