Wind lidar can only directly measure the radial components of wind vectors. So two-dimensional（2D）wind retrieval is crucial for reconstructing wind structure. To address the issue of wind-speed distortion in the 2D wind retrieval using traditional Velocity Azimuth Processing（VAP）algorithm, the relative total variation（RTV）model is incorporated to improve the algorithm. Based on the correlation between wind-speed distortion and radial wind speed, a threshold is established to identify distorted regions ,enabling subsequent correction. Then the RTV model is employed to eliminate irregular textures generated during the correction process and extract the 2D overall wind structure . And local texture features are reconstructed by the texture information from the actual radial wind speed. Experimental results demonstrate that the improved algorithm effectively mitigates the wind-speed distortion issue in VAP algorithm. Compared to the preliminary results retrieved by VAP algorithm, the improved algorithm reduces root mean square error by 0.42 m/s for wind speed and 4.85 for wind direction, significantly improving the accuracy of wind retrieval.