Most of the current research focuses on the improvement and application of spatial and spectral resolution of Hyperspectral Image(HSI). It pays little attention to the comprehensive application of radiation resolution. The radiation resolution reflects the range of the dynamic change of the radiation energy received by the sensor. It detects the small change of the radiation energy of the ground object, which also contains rich ground object information. This study proposes a HSI Radiation Bit Depth Residual Quantization Method to construct Low Bit Depth Hyperspectral Image (LHSI) and Residual Hyperspectral Image (RHSI)with different radiation bit depth levels. Through experiments, LHSI and RHSI of different radiation bit depth levels of HSI and their combinations are used to classify ground objects, and their effects on the classification accuracy of ground objects are analyzed. Experiments show that, based on ensuring a certain classification accuracy, 9-bit LHSI retains the main information of HSI; 4-bit RHSI highlights more details of ground objects than the HSI. The combination of 13-bit LHSI and 3-bit RHSI can not only retain the main information of HSI but also highlight the details of the ground object.