Publications
Detailed Information
Constrained Optimization for Translucent Hindrance Removal from a Single Image : 단일 이미지에서 반투명 방해 요소를 제거하기 위한 제약 최적화
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Jin Young Choi | - |
| dc.contributor.author | Tushar Sandhan | - |
| dc.date.accessioned | 2018-11-12T00:54:52Z | - |
| dc.date.available | 2018-11-12T00:54:52Z | - |
| dc.date.issued | 2018-08 | - |
| dc.identifier.other | 000000152205 | - |
| dc.identifier.uri | https://hdl.handle.net/10371/143042 | - |
| dc.description | 학위논문 (박사)-- 서울대학교 대학원 : 공과대학 전기·컴퓨터공학부, 2018. 8. Jin Young Choi. | - |
| dc.description.abstract | Interstellar, satellite and microscopic imaging methods produce high-resolution images which carry vital information even at a pixel level. Translucent hindrance from reflection, cloud or dusty layer is unavoidable in real-world imaging scenarios. It camouflages the vital image details by altering color and brightness. So the captured image consists of superposition of underlying true object (foreground) and the hindrance (background). In this work we have proposed the optimization model to unravel this superposition from a single input image without losing any image information.
We have identified and formulated three novel translucent hindrance removal tasks namely: eyeglass reflection, high-altitude cloud and microscopic dust removal from a single image. These hindrances seem to be unrelated each other, but after analyzing their characteristics in detail we have managed to relate them statistically. So we were able to develop unified constrained optimization framework, which is flexible enough to handle all translucent hindrance removal tasks and can easily accommodate any new as well as our proposed image priors. Its Fourier transform and look-up table based iterative solution removes hindrances quickly without using any GPU. It can process high-resolution images and produces state-of-the-art results. To the best of our knowledge, the proposed method is the first attempt on formulating and addressing all three important hindrances removal tasks. Our dataset images and source codes are available publicly. This solution has a vital importance in biometrics, satellite and microscopic imaging | - |
| dc.description.tableofcontents | 1 Introduction 1
1.1 Hindrances in computer vision . . . . . . . . . . . . . . . . . . . . . 2 1.1.1 Opaque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.1.2 Translucent . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.1.3 Transparent . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2 Problem formulation . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2.1 Need for hindrance removal . . . . . . . . . . . . . . . . . . 4 1.2.2 Problem statement . . . . . . . . . . . . . . . . . . . . . . . 5 1.3 Motivations and challenges . . . . . . . . . . . . . . . . . . . . . . . 7 1.3.1 Usage of eyeglass reflection removal . . . . . . . . . . . . . . 7 1.3.2 Usage of high-altitude cloud detection and removal . . . . . . 7 1.3.3 Usage of microscopic dust removal . . . . . . . . . . . . . . 8 1.4 Organization of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2 Anti-Glare: Tightly Constrained Optimization for Eyeglass Reflection Removal from a Single Image 9 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.1.1 Use of additional apparatus for reflection removal . . . . . . . 11 2.1.2 Use of multiple images . . . . . . . . . . . . . . . . . . . . . 11 2.1.3 Single image . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.2 Our Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.2.1 Eyeglass Reflection Properties . . . . . . . . . . . . . . . . . 13 2.2.2 Layer Separation Model . . . . . . . . . . . . . . . . . . . . 16 2.2.3 Facial Symmetry Prior . . . . . . . . . . . . . . . . . . . . . 17 2.2.4 Constraint Tightening by Residual Map . . . . . . . . . . . . 18 2.2.5 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.3 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.3.1 Flipping matrix . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.3.2 Illuminations on an eyeglass . . . . . . . . . . . . . . . . . . 24 2.3.3 Dataset details . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3.4 Implementation details . . . . . . . . . . . . . . . . . . . . . 26 2.3.5 Symbiotic Relationship between the Priors . . . . . . . . . . 26 2.3.6 Convergence Analysis . . . . . . . . . . . . . . . . . . . . . 27 2.3.7 Comparison with Previous Methods . . . . . . . . . . . . . . 28 2.3.8 Application: Iris Detection Improvement . . . . . . . . . . . 35 2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3 Removal of High Altitude Clouds, Heterogeneous Fog and Interstellar Clouds from a Single Image 37 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.1.1 Multimodal approaches for cloud removal . . . . . . . . . . . 39 3.1.2 Unimodal approaches . . . . . . . . . . . . . . . . . . . . . . 40 3.1.3 Cloud detection . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.2 Our Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 3.2.1 Aerial Image Gradient Statistics . . . . . . . . . . . . . . . . 42 3.2.2 High Altitude Cloud Properties . . . . . . . . . . . . . . . . 43 3.2.3 Estimating Gradient Statistics for Latent Layer . . . . . . . . 46 3.2.4 Optimization for Latent Layers . . . . . . . . . . . . . . . . . 46 3.3 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 3.3.1 Quantitative Comparison . . . . . . . . . . . . . . . . . . . . 54 3.3.2 Results on Real-World Inputs . . . . . . . . . . . . . . . . . 58 3.3.3 Computation-time . . . . . . . . . . . . . . . . . . . . . . . 59 3.3.4 Optimization Convergence Analysis . . . . . . . . . . . . . . 60 3.4 Heterogeneous fog removal . . . . . . . . . . . . . . . . . . . . . . . 61 3.5 Interstellar cloud removal . . . . . . . . . . . . . . . . . . . . . . . . 62 3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 3.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 4 Separating the Dust from a Single Microscopic Image 64 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 4.2 Microscope and formation of the dusty background . . . . . . . . . . 67 4.2.1 Glass slide and cover slip . . . . . . . . . . . . . . . . . . . . 67 4.2.2 Ubiquitous nature of dust . . . . . . . . . . . . . . . . . . . . 67 4.2.3 Strong attraction of dust towards glass . . . . . . . . . . . . . 67 4.3 Our approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 4.3.1 Translucency of the solid dust particles . . . . . . . . . . . . 68 4.3.2 Generality of our translucent hindrance removal method . . . 70 4.4 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 5 Solving Generalized Non-convex Non-smooth Constrained Optimization Problem 75 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 5.2 Generalized optimization with accommodating multiple priors on background and hindrance layers . . . . . . . . . . . . . . . . . . . . . . 77 5.3 Solving generalized optimization . . . . . . . . . . . . . . . . . . . . 78 5.4 Other algorithms for solving a non-convex, non-smooth and constrained optimization problem . . . . . . . . . . . . . . . . . . . . . . . . . . 80 5.4.1 Optimization solver: GLOBAL . . . . . . . . . . . . . . . . . 81 5.4.2 Multilevel coordinate search: MCS . . . . . . . . . . . . . . . 81 5.4.3 Finite difference and local SQP steps: SQP . . . . . . . . . . 82 5.4.4 Interior-point methods: IP . . . . . . . . . . . . . . . . . . . 82 5.4.5 Genetic algorithms: GA . . . . . . . . . . . . . . . . . . . . . 83 5.4.6 Simulated annealing: SA . . . . . . . . . . . . . . . . . . . . 83 5.5 Solitariness of tightly constrained optimization for eyeglass reflection hindrance removal . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 5.6 Experimental results . . . . . . . . . . . . . . . . . . . . . . . . . . 85 5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 6 Conclusion and future prospects 90 Bibliography 93 Abstract (In Korean) 102 | - |
| dc.language.iso | en | - |
| dc.publisher | 서울대학교 대학원 | - |
| dc.subject.ddc | 621.3 | - |
| dc.title | Constrained Optimization for Translucent Hindrance Removal from a Single Image | - |
| dc.title.alternative | 단일 이미지에서 반투명 방해 요소를 제거하기 위한 제약 최적화 | - |
| dc.type | Thesis | - |
| dc.contributor.AlternativeAuthor | 투샬산드한 | - |
| dc.description.degree | Doctor | - |
| dc.contributor.affiliation | 공과대학 전기·컴퓨터공학부 | - |
| dc.date.awarded | 2018-08 | - |
- Appears in Collections:
Item View & Download Count
Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.