Large-scale particle image velocimetry (LSPIV) has been widely used for measuring the velocity distribution of water surface. The velocity distribution of the area far from the camera tends to have large uncertainty, mainly due to the strong image distortion and reduced image resolution, especially the resolution parallel to the viewing direction. This study proposes the method to overcome this problem, by combining multi-camera LSPIV measurement and combining the result obtained from each camera. Velocity component for the direction orthogonal to the viewing direction (1D velocity component) for each viewing angle is extracted and then combined to reconstruct the 2D velocity field by using the least square algorithm. The result from a rotating disk experiment and the field image is reported.