Real-Time Imaging Algorithm Based on Adaptive Dynamic Tone-Mapping Curves

This paper proposes a tone mapping algorithm based on adaptive dynamic curves for real-time high dynamic range (HDR) image enhancement on FPGA platforms. The method integrates functional branches including logarithmic compression, dark region enhancement, and highlight suppression, while dynamically adjusting the mapping curve according to the intra-frame luminance distribution. Unlike conventional approaches that rely on full-frame luminance statistics, the proposed hardware implementation estimates luminance using an Exponential Moving Average (EMA) based solely on the pixel values input at each clock cycle, and performs the mapping computation directly in the pipeline without buffering the entire frame. This design significantly reduces memory requirements while maintaining the accuracy of luminance adaptation. The algorithm adopts a fixed-point pipelined hardware architecture that relies exclusively on hardware-friendly operations, enabling efficient deployment on FPGA and similar platforms. Experiments conducted on representative scenarios such as darkroom, night street scenes, forest waterfalls, and dim bookshelves demonstrate that the proposed method achieves consistently strong overall performance in both subjective visual quality and multiple objective evaluation metrics, highlighting its effectiveness and practical value for real-time HDR image enhancement applications.