Soil salinization is a critical constraint for agriculture development in arid and semiarid areas, and also one of the most important environmental problems. Therefore, obtaining accurate soil salinization information is crucial to salinization management in those areas. The current study is attempting to derive a relatively straight forward soil salinity index from Landsat TM remote sensing images. First, perform minimum noise fraction (MNF) of the images and calculate their pixel purity index (PPI); select the first three bands that are good to characterize the feature information of a region to construct a MNF spectral eigenspace; then put forward the concept of “vegetation highlight area” by combining field investigations and following the vector space and single line theories and define soil salinization distance index (SDI), so as to enable the multi-dimensional vector space to include the normalized distance from a salinization pixel within the single line to the vegetation highlight area; and in the end verify SDI for precision using the data obtained in field investigations of regions different in salinization level. Results show that in areas low in vegetation coverage, that is, areas moderate and severe in salinization, SDI is more closely related to the average soil salt content in the 0~10 cm soil layer, (R²> 0.83) than in the 0~20 cm soil layer, while in areas high in vegetation coverage, that is, farmland of areas low in soil salinization, it is just the reverse (R²> 0.81), indicating that the overall precision of SDI in prediction of soil salinity in the 0~10 cm soil layer is R²= 0.81, and in the 0~20 cm soil layer R²= 0.72. The findings suggest that the SDI index model is simple and easy- to- construct and yet quite high in precision, so it is of high practical value and can be used to help quantitatively analyze and monitor soil salinization on a large scale in the arid and semiarid areas.