Image contrast is an important factor in distinguishing objects in the image from their background. Low-contrast images, caused by various factors such as poor lighting, are insufficient for human visual perception and many image processing applications. Therefore, image contrast enhancement (ICE) is a necessary preprocessing step in different image processing applications. The main purpose of ICE is to make image objects more easily distinguishable and to improve the quality of visual information in the image. In this paper, image contrast enhancement is studied as an optimization problem. First, a modified version of the gray wolf optimization (GWO) algorithm, a population-based meta-heuristic that mimics the social leadership and hunting behavior of gray wolves in nature, is adapted to the ICE problem. Second, a novel variant of GWO is proposed using Halton low-discrepancy sequence in the population initialization phase, instead of starting completely randomly. Third, unlike previous studies, an effective metric is used as a fitness function to measure the image quality by focusing on the contrast change without using a reference image. The experimental results on the TID2013 and CSIQ datasets show that the proposed Halton sequence-initialized GWO outperforms other variant metaheuristic algorithms and traditional histogram equalization-based methods, according to all utilized various evaluation metrics.