The Distributed Iterative Localization algorithm (DILOC) is an attractive algorithm for wireless sensor network (WSN), which is used to locate sensors (with unknown locations) in m dimensional space. DILOC algorithm needs a minimal number of m +1 anchors (with known their exact locations). However, the convex hull test requires that sensors must be inside of the convex hull, so it will require relatively large communication radius. Considering this defect of DILOC, this paper proposes an improved algorithm that called distributed joint iterative localization (DJIL), through employing loop iteration and weighted average, which is capable of reducing communication radius and enhancing the precision of localization. The DJIL algorithm improves the convex hull test, and uses some sensors, which have known locations, participating in the second localization, and further improving centroid localization algorithm for reliability and practicality. Simulation results show that the communication radius is significantly reduced and error is also decreased for the DJIL algorithm. Compared with DILOC algorithm, if fewer anchors are distributed in region unequally, and the region contains more sensors, DJIL algorithm can reduce the communication radius of 30% and increase location accuracy of 21%. So DJIL algorithm is very flexible for scenarios of anchors uneven distribution.