| dc.contributor | 資訊系 | |
| dc.creator (作者) | 彭彥璁 | |
| dc.creator (作者) | Peng, Yan-Tsung;Cheng, Sheng-Wei;Lin, Yi-Ting;Yu, Shih-Hsien | |
| dc.date (日期) | 2025-04 | |
| dc.date.accessioned | 20-Jan-2026 13:18:55 (UTC+8) | - |
| dc.date.available | 20-Jan-2026 13:18:55 (UTC+8) | - |
| dc.date.issued (上傳時間) | 20-Jan-2026 13:18:55 (UTC+8) | - |
| dc.identifier.uri (URI) | https://ah.lib.nccu.edu.tw/item?item_id=180738 | - |
| dc.description.abstract (摘要) | The non-local means (NLM) algorithm is developed by a non-linear filter that uses spatial and range distances along with each neighbor block patch to perform edge-preserving smoothing of images. For each pixel, a direct computation of NLM requiresO(S2P2 ) operations, whereSis the supported search window radius, andPis the patch radius. In this paper, we propose a novel algorithm to reduce the above operations toO(S2P) based on separable 1-D convolution from 2-D one with mathematical quality lossless property, while the spatial redundancy between two successive image patch windows is made to be applied repeatedly. Experimental results conducted on several image databases demonstrate that the output image quality of our proposed algorithm outperforms that of other state-of-the-art approximation-based NLM algorithms. | |
| dc.format.extent | 101 bytes | - |
| dc.format.mimetype | text/html | - |
| dc.relation (關聯) | IEEE International Conference on Applied System Innovation (ICASI 2025), IEEE International Conference on Applied System Innovation | |
| dc.title (題名) | Quality-lossless time complexity optimization for non-local means | |
| dc.type (資料類型) | conference | |
| dc.identifier.doi (DOI) | 10.1049/icp.2025.2592 | |
| dc.doi.uri (DOI) | https://doi.org/10.1049/icp.2025.2592 | |
