본문 바로가기 주메뉴 바로가기
22대 대한민국 국회 국회정보길라잡이 국회의원검색
회원가입 로그인
HOME

정보검색

소장정보 검색

국회도서관 홈으로 정보검색 소장정보 검색

결과 내 검색

동의어 포함

고급검색

인명/단체명

저자정보 상세정보
인명/단체명을 입력하세요.

키워드

  • 대표어

  • 외국어

-

목차보기

Title Page

Abstract

Contents

Chapter 1. Introduction 11

Chapter 2. Related Work 14

2.1. Single image reflection removal 14

2.2. Diffusion models based on SDE 15

Chapter 3. Preliminary 16

3.1. Stochastic differential equation (SDE) 16

Chapter 4. Proposed Method 17

4.1. Motivation 17

4.2. Method 18

4.2.1. Forward process with SDE 19

4.2.2. Reverse process with SDE 20

Chapter 5. Experimental Results 21

5.1. Experimental Setups 21

5.2. Evaluation with previous methods 21

5.3. Extension experiments 31

5.3.1. Comparison of reflections on glass RG[이미지참조] 31

5.3.2. Comparison with 360-degree image method 32

5.3.3. Comparison with dataset from previous methods 35

Chapter 6. Conclusion 37

References 38

List of Figures

Figure 1.1. (a) Input image and background image without reflection. (b) Background result of reflection removal from the previous method [1] and [2]. (c) The result of background B and reflection R from the... 12

Figure 4.1. This is an overview of the proposed method. The mode can be selected to either restore the background or the reflection scene. On the left is the forward process, which converges the input image... 18

Figure 5.1. The result of qualitative evaluation with state-of-art methods including ERRNet [3], CoRRN[1], IBCLN [4], and Zheng et al [2]. The dataset is postcard in SIR2 [5]. The regions where... 24

Figure 5.2. The result of qualitative evaluation with state-of-art methods including ERRNet [3], CoRRN [1], IBCLN [4], and Zheng et al [2]. The dataset is object in SIR2. The regions where... 25

Figure 5.3. The result of qualitative evaluation with state-of-art methods including ERRNet [3], CoRRN [1], IBCLN [4], and Zheng et al [2]. The dataset is wild in SIR2 [5]. The regions where... 26

Figure 5.4. The result of qualitative evaluation with state-of-art methods including ERRNet [3], CoRRN [1], IBCLN [4], and Zheng et al [2]. The dataset is Zhang et al. The regions where... 27

Figure 5.5. The result of qualitative evaluation with state-of-art methods including ERRNet [3], CoRRN [1], IBCLN [4], and Zheng et al [2]. The dataset is Zhang et al [6]. The regions where... 28

Figure 5.6. The result of qualitative evaluation with state-of-art methods including ERRNet [3], CoRRN [1], IBCLN [4], and Zheng et al [2]. The dataset is Zhang et al [6]. The regions where... 29

Figure 5.7. The qualitative result of reconstructing the scene behind the camera for input image. (a) is the input image and (b) is the groundtruth of the background. (c) shows the restoration of the camera's... 30

Figure 5.8. The results of reflections on the glass with state-of-art methods including ERRNet [3], CoRRN [1], IBCLN [4], and ABS [2]. The top three lines are SIR2 [5], representing postcard, object,... 32

Figure 5.9. The results of reflection removal on the 360-degree images. (a) is the 360-degree image. (b) and (c) is the glass region and reference region which is cropped 256x256. (d) is the result of removing... 33

Figure 5.10. The results of reflection removal on the 360-degree images. (a) is the 360-degree image. (b) and (c) is the glass region and reference region which is cropped 256x256. (d) is the result of removing... 33

Figure 5.11. The results of reflection removal on the 360-degree images. (a) is the 360-degree image. (b) and (c) is the glass region and reference region which is cropped 256x256. (d) is the result of removing... 34

Figure 5.12. The results of reflection removal on the 360-degree images. (a) is the 360-degree image. (b) and (c) is the glass region and reference region which is cropped 256x256. (d) is the result of removing... 34

초록보기

 An image with reflections can be defined as a composition of the background and reflection images, where the two layers are mixed together. Reflection removal is a highly challenging and ill-posed problem that involves separating these two layers. To address this issue, our work utilizes a diffusion model as a type of auto-encoder with stochastic differential equations in a dual manner. This approach allows both the background and reflection images to gradually revert to their original states. Typically, a diffusion model goes through a forward process where the input image is subjected to pure Gaussian noise. However, in reflection removal, where both layers are mixed, the forward process is designed to degrade each image into mixed images. This is achieved by employing stochastic differential equations, enabling the background and reflection images to blend together. To revert to their original states, the values added during the forward process are estimated and subtracted through deep learning. This process facilitates a progressive return to the image of the previous timesteps. Through our approach, we successfully restore both the background and reflection images from a single image containing reflections, allowing for the bidirectional restoration of scenes reflected on light-transmitting material surfaces.

추천서가 (다양한 추천 자료를 만나보세요)

권호기사보기

권호기사 목록 테이블로 기사명, 저자명, 페이지, 원문, 기사목차 순으로 되어있습니다.
기사명 저자명 페이지 원문 기사목차

권호

기사명 원문보기 기사목차
닫기