Visualization of High Dynamic Range Solar Imagery and the Radial Histogram Equalizing Filter

TL;DR

The paper tackles the challenge of visualizing extremely high dynamic range EUV solar imagery by introducing the Radial Histogram Equalizing Filter (RHEF), a parameter-free, radial-annulus percentile remapping method that flattens radial brightness while preserving ordinal contrast. It pairs RHEF with an optional Upsilon redistribution function, a symmetric, dual-sided gamma-like tonal compression, to provide intuitive control over perceptual brightness. RHEF operates on single frames and demonstrates robust, uniform enhancement across on-disk and off-limb regions, with competitive runtimes (e.g., ~1 s for 1024×1024; ~1 min for 4096×4096) and straightforward parallelization. The authors benchmark RHEF against existing methods, show its strengths for morphology-focused analyses (plume detection, loop tracing, synoptic visualization), and provide extensive instrument demonstrations (AIA, SUVI, K-Cor, LASCO C3, PUNCH-simulated data). Implemented in sunkit_image with an optional Python/IDL pathway, RHEF offers immediate improvements for solar visualization and has potential to support real-time monitoring, outreach, and integration with future ML-based feature detection pipelines.

Abstract

Standard visualizations of Extreme Ultraviolet (EUV) solar imagery often fail to convey the full complexity of the Sun's corona, especially in faint off-limb regions. This can leave the misleading impression of the Sun as a bright ball in a dark void, rather than revealing it as the dynamic, structured source of the solar wind and space weather. A variety of enhancement algorithms have been developed to address this challenge, each with its own strengths and tradeoffs. We introduce the Radial Histogram Equalizing Filter (RHEF), a novel hybrid technique that optimizes contrast in high dynamic range solar images. By combining the spatial awareness of radial graded filters with the perceptual benefits of histogram equalization, RHEF reveals faint coronal structures and works out of the box -- without requiring careful parameter tuning or prior dataset characterization. RHEF operates independently on each frame, and it enhances on-disk and off-limb features uniformly across the field of view. For additional control, we also present the Upsilon redistribution function -- a symmetrized cousin of gamma correction -- as an optional post-processing step that provides intuitive programmatic tonal compression. We benchmark RHEF against established methods and offer guidance on filter selection across various applications, with examples from multiple solar instruments provided in an appendix. Implemented and available in both Python sunkit_image and IDL, RHEF enables immediate improvements in solar coronal visualization.

Figures (13)

• Figure 1: Comparison of contrast enhancement techniques applied to the SunPy AIA 171 Å sample image. The top-left panel shows the image with basic gamma correction (i.e., without high dynamic range (HDR) enhancement). The top-right and bottom-left panels show results from the Normalizing Radial Graded Filter (NRGF) and Multiscale Gaussian Normalization (MSGN), respectively (we use these for comparison, but note that they were never designed to be used on-disk). The bottom-right panel displays the result of the Radial Histogram Equalizing Filter (RHEF) and Upsilon $\Upsilon$, which preserves detail across both on-disk and off-limb regions while enhancing the visibility of faint coronal structures.
• Figure 2: Effect of the filter on an AIA 211 image. Each column represents a different filter procedure. The top row shows the image, the middle row shows the full histogram of values as a function of height, and the bottom row shows a histogram of values at one height, as indicated by the red vertical bar in the second row. The blue vertical lines at increasing heights show the limb, detector edge, and optical edge respectively. Note that this figure is showing the effect of $\Upsilon_L \neq \Upsilon_H$. Note also that the "comb-like" structures in the histogram come from the small number of integer values recorded outside the optical path on the detector.
• Figure 3: The $\Upsilon$ redistribution family. Left: curves for different $\Upsilon$ values, with a representative example $\Upsilon=0.35$ in dark blue, alternatives in light blue, and red dashed lines showing standard gamma corrections for reference. Right: recommended default $\Upsilon_L$ and $\Upsilon_H$ values for AIA filters. Symmetric and asymmetric cases are supported.
• Figure 4: The sample 171 filter image from SunPy is displayed in the first panel. This is used to produce an RHEF filtered image, which is then treated with different values for Upsilon $\Upsilon$. Note that the top middle panel shows RHE without Upsilon. Default values are given in Figure \ref{['fig:upsilon_redistribution']} for AIA channels.
• Figure 5: Comparison of enhancement techniques applied to a single AIA 171 Å image. Top row: processed images. Middle row: radial profiles showing the mean and standard deviation of intensity versus distance from disk center. Bottom row: scatter plots showing the full distribution of normalized pixel intensities at each radius. RHEF provides near-uniform radial statistics, while $\Upsilon$ selectively compresses both tails of the intensity distribution. The vertical blue lines indicate the limb, detector edge, and optical edge, respectively.