Utility-Oriented Underwater Image Quality Assessment Based on Transfer Learning

Weiling Chen,Rongfu Lin,Honggang Liao,Tiesong Zhao,Ke Gu,Patrick Le Callet
DOI: https://doi.org/10.48550/arXiv.2205.03574
2022-05-07
Abstract:The widespread image applications have greatly promoted the vision-based tasks, in which the Image Quality Assessment (IQA) technique has become an increasingly significant issue. For user enjoyment in multimedia systems, the IQA exploits image fidelity and aesthetics to characterize user experience; while for other tasks such as popular object recognition, there exists a low correlation between utilities and perceptions. In such cases, the fidelity-based and aesthetics-based IQA methods cannot be directly applied. To address this issue, this paper proposes a utility-oriented IQA in object recognition. In particular, we initialize our research in the scenario of underwater fish detection, which is a critical task that has not yet been perfectly addressed. Based on this task, we build an Underwater Image Utility Database (UIUD) and a learning-based Underwater Image Utility Measure (UIUM). Inspired by the top-down design of fidelity-based IQA, we exploit the deep models of object recognition and transfer their features to our UIUM. Experiments validate that the proposed transfer-learning-based UIUM achieves promising performance in the recognition task. We envision our research provides insights to bridge the researches of IQA and computer vision.
Computer Vision and Pattern Recognition,Image and Video Processing
What problem does this paper attempt to address?
### Problems the paper attempts to solve This paper aims to solve the application problems of image quality assessment (IQA) techniques in specific tasks, especially when there is a low correlation between the utility of an image and its perceived quality (such as fidelity and aesthetics). Specifically: 1. **Limitations of traditional IQA methods**: - Traditional IQA methods mainly focus on the fidelity of an image (i.e., the clarity of image details and textures) and aesthetics (i.e., the harmonious beauty of an image). However, in some specific tasks, such as object recognition, these methods cannot be directly applied because they cannot well reflect the utility of an image in completing the task. 2. **Requirements of specific tasks**: - In practical applications, especially in complex scenarios such as underwater fish detection, the utility of an image is a key factor in determining its quality, not just the fidelity or aesthetics of the image. Therefore, it is necessary to develop a utility - based IQA method to evaluate the performance of an image in specific tasks. 3. **Research background**: - Underwater images are very different from natural images due to their special statistical properties, visual features, and distortion types. Most of the existing IQA methods are for natural images or videos and cannot be directly applied to the quality assessment of underwater images. 4. **Research objectives**: - To solve the above problems, this paper proposes a utility - oriented IQA method based on transfer learning, which is especially suitable for underwater fish detection tasks. This method constructs an underwater image utility database (UIUD) and a learning - based underwater image utility metric (UIUM) to evaluate the utility of an image by using the features of a deep model. ### Specific problem descriptions - **How to evaluate the utility of underwater images in fish detection tasks**: Traditional IQA methods cannot be directly applied to underwater fish detection tasks because they mainly focus on the fidelity and aesthetics of images and ignore the utility of images in completing specific tasks. - **Lack of a database suitable for underwater image utility assessment**: Most of the existing IQA databases are for natural images and lack large - scale databases specifically for underwater image utility assessment. - **How to use transfer learning to improve the performance of IQA models**: Due to the limited number of training samples in existing databases, deep - learning - based IQA methods are prone to overfitting. Therefore, it is necessary to use transfer learning techniques to extract useful features from related source tasks to improve the performance of IQA models. By solving these problems, the method proposed in this paper can more accurately evaluate the utility of images in underwater fish detection tasks, thus providing a valuable reference for subsequent research and applications.