With the advancement of remote sensing technology, the mineralization-related alteration anomalies can now be used as independent indicators for mineral exploration. Crosta technique, as a method for mineralization and alteration information extraction, played an important role in mineral exploration. However, the traditional Crosta technique calculates the eigenveetors of each component based on the statistical analysis of the entire image. It will produce a lot of noise in the abnormal component and reduce the accuracy of the alteration information extracted. This paper presents an improved method based on local variable window. In the improved method, the entire image is divided into a number of independent statistical analysis units by the local variable window. In each local variable window, principal component analysis and judgment of the abnormal component will comply after the elimination of interference information such as water, clouds, snow and so on. The improved method can reduce the background noise, filter interference information, and identify weak alteration information effectively. Using the improved Crosta technique, mineralization-related alteration anomalies are extracted to guide mineral resources prospecting in the Mohailaheng area, Qinghai province. The NWW-trending faults are the principal faults in the Mohailaheng area, which control the distribution of the strata and the mineralized zones. The combinations of silieification, ferritization and carbonatization anomalies within the fault zones are useful indicators for mineral exploration. There are some intensive alteration anomalies of silicifieation and ferritization in the intersections of fault F3, F4 and F5, which will be the high priority targets for Pb-Zn deposit exploration.