상세 보기
TPRNet: Texture PReserving Network for Realistic Scenery Image Extrapolation
- Lim, Hyoung Jun;
- Lee, Jooyoung;
- Choi, Jongwook;
- Park, Soo Hyun;
- Choi, Jongwon
WEB OF SCIENCE
0SCOPUS
0초록
We propose a Texture PReserving Network (TPRNet) allowing for high-quality extrapolation results that maintain the original image’s clarity and structure, even with diverse resolutions, without the need for additional re-training. For this purpose, we employ a novel approach by reusing patches from the input image for extrapolation, ensuring clarity and contextual consistency. The rearrangement of patches is guided by graph-based features, leveraging both contextual and structural information to select the most appropriate patches for each location. Furthermore, we address possible challenges in patch-based extrapolation, such as unnatural patch borders and repetitive patterns, through a diffusion-based refinement module. The outstanding performance of our framework is validated through qualitative comparisons and comprehensive quantitative analysis, further presenting its suitability for practical applications.
- 제목
- TPRNet: Texture PReserving Network for Realistic Scenery Image Extrapolation
- 저자
- Lim, Hyoung Jun; Lee, Jooyoung; Choi, Jongwook; Park, Soo Hyun; Choi, Jongwon
- 발행일
- 2027
- 유형
- Proceedings Paper
- 권
- 16825
- 페이지
- 639 ~ 653
- 언어
- ENG
- 출판사
- Springer Science and Business Media Deutschland GmbH
- 발행국가
- 미국
- 분량
- 15 페이지
- ISSN
- E 1611-3349
P 0302-9743