When the technology of the image measurement is applied in the double secondary permanent magnet linear motor rotor position measurement, the selection of the target source image is very important for its measurement precision and anti-interference . According to the width of the standard deviation and the gray gradient sum to construct different aperiodic fence image, the original picture surface is used. The quadtree decomposition method is used to decompose the collected fence image into each piece. According to the edge texture of the block image and its constraint conditions of neighborhood pixels to distinguish as well as eliminate the influence of edge texture of noise, the noise levels of block image are achieved accurately. Each image block is weighted, the whole image with noise results from the average level. Finally, peak signal to noise ratio is used to evaluate the quality of the image, and the phase correlation algorithm is used to verify its anti-interference. Results show that the gray aperiodic fence image's peak signal to noise ratio is better and its similarity peak is higher for the width standard deviation change in ［8.1,8.3］ and thegray gradient sum change in ［18 000,22 000］, and its anti-interference is better and can realize high precision measurement of the linear electric motor rotor position.