Beneath the Surface: The Role of Underwater Image Enhancement in Object Detection

TL;DR

This work investigates how underwater image enhancement (UIE) affects underwater object detection (UOD) by evaluating nine UIE models across two challenging datasets and examining detection performance at the image level rather than the dataset level. It introduces a joint quantitative-qualitative framework and a Q-index to study quality distributions, revealing that enhancement often degrades dataset-level detection but can improve per-image detection for specific cases, particularly for low-quality images. The study finds no reliable correlation between traditional image quality metrics and detection performance, underscoring the need for metrics that jointly account for visual quality and detectability. A key takeaway is that selective enhancement—applied to individual images based on expected detection gains—can improve overall detection, with a mixed set of original and enhanced images yielding the best per-image mAP.

Abstract

Underwater imagery often suffers from severe degradation resulting in low visual quality and reduced object detection performance. This work aims to evaluate state-of-the-art image enhancement models, investigate their effects on underwater object detection, and explore their potential to improve detection performance. To this end, we apply nine recent underwater image enhancement models, covering physical, non-physical and learning-based categories, to two recent underwater image datasets. Following this, we conduct joint qualitative and quantitative analyses on the original and enhanced images, revealing the discrepancy between the two analyses, and analyzing changes in the quality distribution of the images after enhancement. We then train three recent object detection models on the original datasets, selecting the best-performing detector for further analysis. This detector is subsequently re-trained on the enhanced datasets to evaluate changes in detection performance, highlighting the adverse effect of enhancement on detection performance at the dataset level. Next, we perform a correlation study to examine the relationship between various enhancement metrics and the mean Average Precision (mAP). Finally, we conduct an image-level analysis that reveals images of improved detection performance after enhancement. The findings of this study demonstrate the potential of image enhancement to improve detection performance and provide valuable insights for researchers to further explore the effects of enhancement on detection at the individual image level, rather than at the dataset level. This could enable the selective application of enhancement for improved detection. The data generated, code developed, and supplementary materials are publicly available at: https://github.com/RSSL-MTU/Enhancement-Detection-Analysis.

Figures (11)

• Figure 1: On the left, The quality distribution of the Original images of the CUPDD dataset saleem2023multi based on the Q-index. On the right, the distributions of the change in quality after enhancement by different models.
• Figure 2: On the left, The quality distribution of the Original images of the RUOD dataset fu2023rethinking based on the Q-index. On the right, the distributions of the change in quality after enhancement by different models.
• Figure 3: Randomly selected Original images from each available quality bin of CUPDD dataset. The corresponding Q-index values are color-mapped and placed under each image.
• Figure 4: Randomly selected Original images from each available quality bin of RUOD dataset. The corresponding Q-index values are color-mapped and placed under each image.
• Figure 5: Inference visualization of the Original and domain detectors of five random images on the CUPDD dataset.