A Modified Retinex Algorithm for Visualization of High Dynamic Range Images

To solve the problem of high dynamic range (HDR) image visualization in consumer electronics products, this paper proposes an improved Retinex algorithm. The proposed algorithm has three aspects of improving the traditional center/surround Retinex algorithm: First, fast bilateral filter is used instead of Gaussian filter, which not only avoids the halo artifact, but also improves the computing speed. Second, a semi-automatic gain/offset method independent of color channels is developed, which can automatically calculate the position of the clip off in the histogram based on the image content, and effectively improve the contrast of the final image. Finally, unique user parameters can be obtained based on the integration of a large number of parameters in the traditional Retinex, and this parameter allows the user to achieve a balance between computational speed and image quality. Compared with other types of HDR image visualization algorithms, the proposed algorithm has excellent visibility detail and color fidelity. It can be used in consumer digital cameras, monitors, and image post-rendering software.