{
  "annotation_id": "35510cbb-b604-41e6-842b-092d2942664f",
  "date_created": "2026-02-17T05:53:20.336000Z",
  "date_modified": "2026-02-17T05:53:20.336000Z",
  "file_hash": "c40833a5c53567a86901907f7d6b51a8bd1c9fcf16a2b02815df538afd8c33c0",
  "private": false,
  "record": {
    "abstract": "With advances in artificial intelligence, image processing has gained significant interest. Image super-resolution is a vital technology closely related to real-world applications, as it enhances the quality of existing images. Since enhancing fine details is crucial for the super-resolution task, pixels that contribute to high-frequency information should be emphasized. This paper proposes two methods to enhance high-frequency details in super-resolution images: a Laplacian pyramid-based detail loss and a repeated upscaling and downscaling process. Total loss with our detail loss guides a model by separately generating and controlling super-resolution and detail images. This approach allows the model to focus more effectively on high-frequency components, resulting in improved super-resolution images. Additionally, repeated upscaling and downscaling amplify the effectiveness of the detail loss by extracting diverse information from multiple low-resolution features. We conduct two types of experiments. First, we design a CNN-based model incorporating our methods. This model achieves state-of-the-art results, surpassing all currently available CNN-based and even some attention-based models. Second, we apply our methods to existing attention-based models on a small scale. In all our experiments, attention-based models adding our detail loss show improvements compared to the originals. These results demonstrate our approaches effectively enhance super-resolution images across different model structures.",
    "arxiv_id": "2601.09410",
    "authors": [
      "Sangjun Han",
      "Youngmi Hur"
    ],
    "categories": [
      "cs.CV"
    ],
    "doi": "10.1049/ipr2.70238",
    "journal_ref": "IET Image Processing, 2025; 19:e70238",
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "title": "Detail Loss in Super-Resolution Models Based on the Laplacian Pyramid and Repeated Upscaling and Downscaling Process",
    "url": "https://arxiv.org/abs/2601.09410",
    "version": "v1"
  },
  "schema_id": "dorsal/arxiv",
  "source": {
    "execution_id": "3f5b937d-fca1-43f9-9000-596a42c3abd8",
    "id": "arXiv Dataset",
    "type": "Model",
    "variant": "snapshot-2026-01-17",
    "version": "0.1.0"
  },
  "user_id": 1000002
}