The use of traditional algorithms cannot obtain a well-focused and cross scaled ISAR image for non-uniform rotating targets. To solve this problem, a method is proposed to estimate the rotating parameters when targets undergo non-uniform rotation. Firstly, the ratio between the angular acceleration and the angular velocity is estimated by one dimension search. By non-uniformly sampling the raw profile sequence according to the estimated ratio, the phase error caused by angular acceleration can be compensated and a focused ISAR image can be obtained. Based on the rule of image quality, the rotating velocity of the resample signal can be estimated by another one dimension search. Furthermore, the scalar of cross-range can be obtained. A focused, cross-range scaled ISAR imaging can be obtained for non-uniform rotating targets based the proposed algorithm. Experimentations with the simulated and measured ISAR data demonstrate the effectiveness of the algorithm.