A detecting method for motor rotor position was proposed based on image measurement technique. Firstly, the image of motor rotor was taken by a high-speed area CCD ceaselessly when the motor was rotating; then every two collected adjacent images were converted from a Descartes coordinate to a polar coordinate; finally, these images in the polar coordinate were calculated by phase-correlation algorithm, and the rotation angle was obtained according to the position of the phase-correlation peak. As the texture information of calculated image was related to the accuracy of phase-correlation, the calculated image was reconstructed to obtain an image to assure the right calculated result for implementing exact detection. Meanwhile, detection efficiency was required to increase, so the adaptive method was used in polar transformation, namely, only part of the pixels in the image was sampled according to the need of measurement precision. Experimental results show that the measurement accuracy of the proposed method is 100% and this measurement can be achieved in the environment with some different illumination intensities. Moreover, the calculating time can be reduced by over 50% as compared with that using polar transformation directly. The proposed method provides a reliable way for the detection of motor rotors.