An Image Quality Evaluation and Masking Algorithm Based On Pre-trained Deep Neural Networks

TL;DR

The paper tackles the challenge of automated image quality assessment and masking in large astronomical datasets. It introduces a reduced-reference approach using a convolutional autoencoder trained on high-quality images to learn a feature space, enabling reconstruction-based quality scoring and patch-level masking via reconstruction discrepancies. The framework is evaluated on simulated data with varying PSF FWHM, complex background noise, and real GWAC observations, showing robust PSF sensitivity and effective noise masking that improves photometric accuracy and reduces false alarms. This method can accelerate data processing pipelines in sky surveys and time-domain astronomy by reducing the need for human intervention and enabling targeted processing of contaminated regions.

Abstract

With the growing amount of astronomical data, there is an increasing need for automated data processing pipelines, which can extract scientific information from observation data without human interventions. A critical aspect of these pipelines is the image quality evaluation and masking algorithm, which evaluates image qualities based on various factors such as cloud coverage, sky brightness, scattering light from the optical system, point spread function size and shape, and read-out noise. Occasionally, the algorithm requires masking of areas severely affected by noise. However, the algorithm often necessitates significant human interventions, reducing data processing efficiency. In this study, we present a deep learning based image quality evaluation algorithm that uses an autoencoder to learn features of high quality astronomical images. The trained autoencoder enables automatic evaluation of image quality and masking of noise affected areas. We have evaluated the performance of our algorithm using two test cases: images with point spread functions of varying full width half magnitude, and images with complex backgrounds. In the first scenario, our algorithm could effectively identify variations of the point spread functions, which can provide valuable reference information for photometry. In the second scenario, our method could successfully mask regions affected by complex regions, which could significantly increase the photometry accuracy. Our algorithm can be employed to automatically evaluate image quality obtained by different sky surveying projects, further increasing the speed and robustness of data processing pipelines.

Figures (11)

• Figure 1: The figure illustrates the structure of the autoencoder, which first encodes input images into latent vectors and subsequently decodes the latent vectors to output images.
• Figure 2: The figure displays the structure of the neural network proposed in this paper, which incorporates multiple convolutional neural networks and residual blocks to efficiently encode the features of images.
• Figure 3: The figures presented here depict simulated images generated based on the previously defined parameters. As illustrated, these images encompass a variety of celestial objects, including stars and galaxies.
• Figure 4: This figure shows MAE of new sets of simulated images obtained by our methods (from FWHM = 0.5 to FWHM = 2.0). The black straight line stands for mean values of all images and the red dotted line stands for MAEs of the same image with different blur levels.
• Figure 5: The figure shows simulated images with different levels of complex background noise. The second and third parts show the results of the same original simulated image using different levels of complex background noise. The first column image represents the original simulated image, while the second to fourth columns images represent images with different complex background noise. Images of columns 5th to 7th are results obtained by our method with different mask sizes and combinations of original image and the contour of ground-truth complex background noise distribution.