<p>Image segmentation is the basis of image processing for image recognition and image analysis. Threshold segmentation is a common method in image segmentation, and Tsallis relative entropy is one of the important threshold segmentation methods. However, as the threshold value increases, the threshold is difficult to determine. Heuristic algorithms have certain advantages in solving such problems as they can effectively address the challenge of obtaining high-quality thresholds in Tsallis relative entropy partitioning. This study proposes a Novel Pyramid Model Multi-Strategy Grey Wolf Optimization (NPMGWO) algorithm to address the challenges associated with obtaining high-quality thresholds in Tsallis relative entropy-based image segmentation. The NPMGWO introduces a novel pyramid model structure that enhances the Grey Wolf Optimization (GWO) by providing a large number of reference samples to support its population. The most important innovation introduced by NPMGWO is the integration of signal-to-noise ratio. With the introduction of dynamic coordinated learning strategy, the signal-to-noise ratio can make the updating of grey wolf individuals' positions more intelligent and rational. Additionally, a novel computational method of nonlinear convergence factor is introduced to improve the convergence ability of the algorithm, thereby enhancing its ability to find optimal solutions. In the experiments of Tsallis relative entropy threshold segmentation, NPMGWO was compared with various algorithms and their variants. The findings clearly demonstrate that NPMGWO excels in terms of segmentation quality and stability, offering a superior performance in addressing such challenges. In conclusion, the NPMGWO algorithm represents a significant advancement in the field of image segmentation.</p>