Ambisonics is a powerful technique for capturing and reproducing spatial sound within the spherical harmonic domain. While low-order Ambisonics (B-format) works well at lower frequencies, its spatial resolution diminishes as frequencies increase. In contrast, higher-order Ambisonics (HOA) offers greater spatial accuracy but requires more complex hardware and increased storage. To overcome these limitations, this study investigates upscaling B-format signals to HOA using the Regularized Multiple Mearurement FOCal Underdetermined System Solver (RM-FOCUSS) approach. Regularization plays a pivotal role in addressing the sparse inverse problem, preventing overfitting, and enhancing the robustness and efficiency of the method compared to traditional sequential forward techniques. By incorporating regularization, the proposed technique significantly improves the upscaling process by minimizing diversity measures. The method’s performance is assessed through mean square error metrics and validated via subjective listening tests, showing superior spatial resolution and overall performance compared to existing state-of-the-art methods.