Neutral wind plays a critical role in the dynamics of the upper atmosphere, and accurate measurements of thermospheric wind fields are essential for the comprehensive understanding and modeling of the ionosphere-thermosphere (IT) system. The Doppler asymmetric spatial heterodyne (DASH) interferometer employs a limb-viewing observational technique, similar to the occultation method, to measure atmospheric wind fields at a fixed viewing angle along the Earth’s tangent direction. In practical applications, raw interferograms are affected by both background and instrumental noise, which reduces fringe contrast. In the frequency domain, this degradation is reflected in weakened dominant frequency components and a low spectral signal-to-noise ratio (SNR), potentially resulting in biased frequency estimation or main-lobe energy leakage. Enhancing fringe contrast is therefore essential for amplifying the dominant frequency peak and improving the accuracy of phase retrieval. Consequently, the development of an effective contrast enhancement strategy is vital for enhancing wind speed measurement performance, particularly under low-signal conditions where it holds substantial practical value.

During the inversion process of low-contrast interferograms, the spatial frequency of interference fringes is highly sensitive to preprocessing due to the low SNR, which can easily cause phase discontinuities and significantly reduce the accuracy of wind field inversion. To address this challenge, we present a contrast-adaptive enhancement algorithm for spaceborne DASH interferograms. As illustrated in Fig. 4, the algorithm first analyzes the image grayscale histogram and adaptively determines the clipping ratio by combining the local standard deviation, histogram dynamic range, and entropy. This ratio, along with the cumulative distribution function, defines the clipping region, effectively mitigating the influence of outlier pixels on contrast enhancement. The clipped grayscale range is then linearly mapped to the full dynamic range, resulting in a substantial improvement in interferogram contrast. The enhanced image is then processed through interferogram preprocessing and wind field inversion, enabling stepwise analysis of atmospheric wind speed distributions at different altitudes.

Despite ongoing advancements, existing interferogram contrast enhancement techniques remain limited in effectiveness. Conventional approaches such as linear stretching and adaptive histogram equalization (AHE) offer only moderate performance under challenging conditions. In this paper, simulated DASH interferograms with varying levels of contrast and superimposed random noise are processed using four methods: the proposed adaptive threshold regulation algorithm, frequency-domain enhancement, contrast-limited adaptive histogram equalization (CLAHE), and standard AHE. As shown in Fig. 6, the first curve represents the output of the proposed method. Under low-contrast conditions, this method exhibits superior noise suppression compared to the other methods, effectively enhancing fringe visibility while minimizing phase fluctuations. As illustrated in Fig. 7, results averaged across multiple trials reveal that when interferogram contrast exceeds 0.4, additional enhancement leads to a decline in inversion accuracy. This is primarily due to the distortion of fringe structure, which introduces spurious frequency components and shifts the dominant frequency in the Fourier domain, thereby impairing the precision of phase extraction. Therefore, for interferograms with inherently high contrast, it is advisable to bypass contrast enhancement and directly apply frequency-domain analysis to preserve the integrity of phase information. To further assess the relationship between fringe contrast and wind retrieval uncertainty, an indoor wind simulation experiment is conducted using the DASH optical system. As depicted in Fig. 9, when the interferogram contrast falls below 0.2, the signal becomes weak, and fringes are highly susceptible to noise, resulting in substantial errors in retrieved wind speeds. Following contrast enhancement, the retrieved wind speeds ranged from 22 to 28 m/s, with corresponding errors between 2.6~8.6 m/s. The primary contributors to these errors include uncertainties in the rotational speed of the wind-generating disc and deviations in the optical path. In addition, wind field retrievals are conducted on three sets of satellite limb-viewing data acquired on June 16. As shown in Fig. 11(a), the inversion results from the unenhanced interferogram closely follow the trend predicted by the horizontal wind model 2014 (HWM 14), yielding a root-mean-square (RMS) error of 9.2 m/s. After applying the proposed contrast enhancement [Figs. 11(b) and (c)], the SNR of the interferograms improved by more than a factor of 3, and the accuracy of the retrieved wind fields increased markedly. These RMS errors decreased from 643.55 m/s and 541.71 m/s to 17.99 m/s and 19.62 m/s, respectively, corresponding to error reductions of 97.2% and 96.4%, with an average reduction of 96.8%.

Through simulation and experimental analyses, we confirm that low-contrast interferograms adversely affect wind speed retrieval accuracy. Additionally, significant variations exist among different contrast enhancement algorithms in their ability to recover interferogram details. These results demonstrate that the contrast enhancement method based on adaptive threshold regulation effectively improves phase retrieval accuracy for low-contrast interference fringes. However, for high-contrast fringes, reconstructing the grayscale histogram can lead to a loss of phase information, and thus, no further contrast enhancement is applied following the optimized enhancement strategy. Based on experimental results, the error is controlled within 10 m/s in both simulation and indoor experiments, while the variance between satellite-based data and the HWM 14 remains within 20 m/s, reflecting a small yet statistically significant difference. Part of this discrepancy can be attributed to instrumental errors, such as noise interference and zero-wind calibration. The remaining differences may arise from atmospheric wind fluctuations or mismatches between the observation geometry and the HWM 14’s projection.