This paper proposes a novel distance calibration and gray-level image generation method for the SR4500 Time of Flight (ToF) camera. For the distance measurement, the Gaussian Process Regression (GPR) model is proposed to calibrate the range error. With this model, fewer parameters need to be estimated and the model complexity is reduced. In order to eliminate the error data caused by the object edge, the distance threshold which determined by the range data of the current pixel is selected to remove the mixed pixel. To acquire gray-level images from the amplitude data, the amplitude histogram is used to select the amplitude threshold of the reflected light, and grayscale images can be generated through normalizing the amplitude value with this threshold. The experiment results illustrate the feasibility and effectiveness of the proposed algorithms.