The differential synthetic aperture radar interferometry technique of permanent scatterers (PS-DINSAR) is an important method in the phase analysis of synthetic aperture radar data. Selection of common master image in PS-DINSAR widely influences the data processing results. Most of existing selection methods of common master image are based the statistical characteristics on the baseline of the interferogram and thus cannot ensure best selected results. A method based on fixed weights of observed values in the surveying adjustment was used in this study to extract an optimum image. The temporal baselines, effective spatial baselines, and Doppler centroid frequency difference of image pairs were incorporated in this technique. The sum of normalized weights model was utilized based on integrated correlation coefficient algorithm and minimum sum of three baseline algorithm to obtain optimal common master image. Then, the procedure of selecting common master images based on the idea of maximum sum of normalized weights was introduced and used to test the selection of common master image by using 19 images of ERS-1/2. Result shows that, in comparison with data generated separately by integrated correlation coefficient algorithm and minimum sum of three baseline algorithm, the outcome produced by maximum sum of normalized weights model in this study exhibits considerably better statistical property in temporal baselines, effective spatial baselines, and Doppler centroid frequency difference of image pairs. Moreover, the interferograms show that the maximum, average, and standard deviation values are less than those of others in temporal baselines, effective spatial baselines, and Doppler centroid frequency difference. Therefore, the selected common master is robust and stable. This study provides an effective method for the optimization of common master image and presents certain guiding significance for selecting common master image.