Silicon-based sensor cameras are known to be sensitive in the near infrared spectral range, in which a small temperature variation leads to a large modification in the image gray level. It induces acquired images with local saturation or poor dynamic range of gray levels. In order to address this problem, the present study proposes an innovative technique to precisely and automatically adjust the exposure time to obtain stable gray level images when the temperature evolution occurs on the surface of the observed object. Two algorithms, including linear algorithm and Planck’s algorithm, are proposed to predict the exposure time to obtain stable gray level images. Blackbody heating experiment is conducted to validate the accuracy of these two algorithms, and the result indicates that stable gray level images can be obtained using Planck’s algorithm. Moreover, this technique is applied to the specimen heating experiment, and the stable gray level images can also be obtained using Planck’s algorithm. These two experimental results prove that the technique is effective and reliable. Finally, the thermal fields are reconstructed on images of blackbody.