In this chapter, an alternative image segmentation method is proposed to find the multi-level thresholding values. The proposed method is based on the tree seed algorithm (TSA) which emulates the tree generations. The proposed TSA used the maximum between class variance criterion (Otsu) as a fitness function. In order to evaluate the performance of the proposed method a set of images are used and the results and compared with other methods. The experimental results illustrated that the proposed TSA method has better performance to determine the multi-level thresholding problem for image segmentation than other methods regarding performance measures.