Abstract: Regarding the problems of low resolution of conventional logging data which can’t effectively reflect grain size of sandy conglomerate and taking the sandy conglomerate reservoir of Baikouquan formation in Mahu sag, Junggar basin as an example, the paper proposes a method to automatically recognize lithologies of complex sandy conglomerate reservoirs by using micro-resistivity imaging logging. Firstly, according to the differences in grain size, the sandy conglomerate can be classified into mudstone, sandstone, fine conglomerate, small medium-sized conglomerate and big medium-sized conglomerate. Based on the core calibrating logs, the characteristics of the micro-resistivity imaging logging are analyzed and summarized for different lithologies. With the technology of image processing, 5 characteristic values such as contrast, relevancy, entropy, uniformity and energy of the sandy conglomerate samples are calculated through image gray scale transformation and gray scale co-occurrence matrix establishment. On the basis of multivariate statistics, Bayes discriminant method is used to build lithology discriminant functions for mudstone, sandstone, fine conglomerate, small medium-sized conglomerate and big medium-sized conglomerate and the lithologies of the rocks can be determined. The results show that the lithologies identified by the micro-resistivity imaging logging highly match with those obtained from core analysis and good identification effects have been gained