Abstract: To address the problems of insufficient temperature measurement accuracy and low resolution of current thermal imaging cameras, a temperature super-resolution model with an enhanced detail high-frequency component is developed by integrating a high-frequency filter block. The model first extracts the shallow features of a feature map through a convolutional layer. Second, a high-frequency filter block is introduced to highlight the high-frequency information, and jump joins are used to combine the raw data with high-frequency information. Finally, the temperature data are upsampled via convolution and pixel rearrangement, thus improving the resolution. This experiment is conducted on a self-constructed dataset, and the experimental results show that the enhanced detail high-frequency component model outperforms the fast super-resolution convolutional neural network and enhanced deep super-resolution network models.