FreGS is a novel 3D Gaussian splatting method that addresses the over-reconstruction issue in 3D Gaussian splatting by introducing progressive frequency regularization. The method leverages frequency space to improve Gaussian densification and novel view synthesis. FreGS performs coarse-to-fine Gaussian densification by progressively regularizing low-to-high frequency components using a frequency annealing technique. This approach minimizes the discrepancy between the frequency spectrum of rendered images and the corresponding ground truth, leading to high-quality Gaussian densification and reduced blur and artifacts in rendered images. Experiments on multiple benchmarks, including Mip-NeRF360, Tanks-and-Temples, and Deep Blending, show that FreGS achieves superior novel view synthesis and outperforms the state-of-the-art consistently. The method is effective in mitigating over-reconstruction by guiding Gaussian densification through frequency regularization, which complements pixel-level losses in the spatial domain. FreGS improves the quality of rendered images with less blur and more fine details, demonstrating its effectiveness in enhancing 3D Gaussian splatting. The proposed method is validated through extensive experiments and visualizations, showing improved Gaussian densification and novel view synthesis compared to existing methods.FreGS is a novel 3D Gaussian splatting method that addresses the over-reconstruction issue in 3D Gaussian splatting by introducing progressive frequency regularization. The method leverages frequency space to improve Gaussian densification and novel view synthesis. FreGS performs coarse-to-fine Gaussian densification by progressively regularizing low-to-high frequency components using a frequency annealing technique. This approach minimizes the discrepancy between the frequency spectrum of rendered images and the corresponding ground truth, leading to high-quality Gaussian densification and reduced blur and artifacts in rendered images. Experiments on multiple benchmarks, including Mip-NeRF360, Tanks-and-Temples, and Deep Blending, show that FreGS achieves superior novel view synthesis and outperforms the state-of-the-art consistently. The method is effective in mitigating over-reconstruction by guiding Gaussian densification through frequency regularization, which complements pixel-level losses in the spatial domain. FreGS improves the quality of rendered images with less blur and more fine details, demonstrating its effectiveness in enhancing 3D Gaussian splatting. The proposed method is validated through extensive experiments and visualizations, showing improved Gaussian densification and novel view synthesis compared to existing methods.