A combined method for histogram equalization of high dynamic range images

When working with raw images obtained directly from the equipment matrix, specific problems arise associated with a large dynamic range. The paper proposes a combined histogram correction method that can significantly improve the contrast of such raw images with a large dynamic range. In the combined method, soft clipping of highlights in the histogram is performed using a clustering algorithm based on partitioning the feature space and gamma correction of the clipped area. The clustering algorithm used manages to identify the cutoff point, both in the presence and absence of highlights in the image. The method also produces weak edge accentuation based on Sobel filters. To improve the histogram, the well-known Contrast Limited Adaptive Histogram Equalization method is used. In this case, a combination of transformations with different mesh sizes is used, which allows one to achieve much better results than when selecting one optimal transformation. These algorithms are described in detail and illustrations for comparison are provided.