In this paper, an efficient three-stage codebook index search algorithm is proposed for quantized equal gain transmission (QEGT) based closed-loop multiple-input multiple-output systems. At the 1st stage, the codewords are grouped into a small number of categories by utilizing Grassmannian criterion or Lloyd's algorithm. At the 2nd stage, the optimal centroid is selected within those of the groups, which maximizes the capacity criterion. At the 3rd stage, the optimal index of QEGT codebook is determined within the group of the selected centroid. Monte-Carlo simulations are presented to show that the codebook index search complexity is halved whilst maintaining the almost same symbol error rate.