Image geometric distortion correction is an important content of digital image processing. It is necessary for the image with geometric distortion and distortion to eliminate these distortions before being carried on detection and data analysis. The correction of the trapezoidal distortion has a higher practical value in the reality using, such as remote sensing image processing, license plate detection, 3D flight simulation module. These have applied in the real life and the training activities. The graduates design takes Visual C++6.0 as the working platform through the compiling procedure to achieve the corresponding algorithm. The treating processes have supposed an original map picture in advance. This picture can obtain the seat punctuation and the gray value of the current picture through the degenerated algorithm, and return to the gray value to the corresponding seat punctuation of the picture supposed in advance. Thus we can realize the picture return to original state. The procedure used the nearest neighbor and bilinear interpolation methods to compute gray value. It has confirmed accuracy and the feasibility of the trapezoid degeneration and the adjustment model. Through analyzing the test, we can know that the nearest neighbor method of correction calculation is faster and less complex than bilinear interpolation. But bilinear interpolation has batter quality than nearest neighbor method after the image having been corrected.