ObjectiveThe cold atomic absorption imaging technique converts the spatial distribution of ultracold erbium atoms into digital images to extract spatial distribution features, temperature, density, and other information by comparing the light intensity distribution with and without atoms, so a better imaging quality is important for studying the long-range dipole-dipole interaction of ultracold erbium atoms. In practical measurements, the light interference and mechanical vibrations of the imaging system usually cause fringe patterns, limiting the accuracy of imaging and the ability to extract physical parameters. In addition to hardware measures such as using an excellent anti-reflection coating and optimizing the optical path to reduce interference, utilizing algorithms for fringe denoising possesses broader reliability and practicality.MethodThis paper utilizes Principal Component Analysis (PCA) as a powerful statistical tool for analyzing large experimental datasets, specifically for fringe removal in absorption imaging of ultracold erbium atoms. Our approach incorporates spatial shifting of the image to broaden the basis, effectively addressing the common mismatches between absorption and reference images due to mechanical vibrations within the imaging system. Masking operations were employed to eliminate the influence of atomic clouds. Ultimately, an ideal reference image with the same fringe patterns for the absorption image was reconstructed to suppress the fringe patterns caused by light field difference. In order to facilitate operations, we have designed and developed software based on PCA for ultracold erbium atom absorption imaging using the Matlab App Designer.Results and DiscussionsThis study demonstrates an algorithm that efficiently removes fringe patterns in absorption imaging of ultracold erbium atoms. By utilizing this algorithm, we have successfully reduced the noise level to approximately $1/2$ of the theoretical limit achieved by traditional methods. Additionally, our approach corrects pixel value differences arising from fluctuations in probe beam power over time, improving the accuracy of extracted temperature and spatial distribution information.ConclusionThis paper applies a denoising algorithm based on Principal Component Analysis to address the fringe noise issue in absorption imaging of ultracold erbium atoms. By using principal component analysis to decompose the extended reference image set and reconstruct an ideal reference image with the same fringe patterns for the absorption image, we effectively suppress the unwanted fringe patterns in the atomic spatial distribution. The quantitative analysis shows that this method suppresses the noise to the quantum limit shot noise level, and the extraction of the ultracold atomic temperature is more accurate after denoising. Based on this, a denoising software for ultracold erbium atom absorption imaging was designed and developed using the Matlab App Designer.