Robust Image Synthesis with Adaptive N D Tensor Voting
This paper presents a method for repairing severely damaged images of natural scenes by using an adaptive N D Tensor Voting algorithm. The approach handles the mixture of textures and colors, inhomogeneity of patterns, and irregular object shapes commonly found in real-world images.
- Uploaded on | 0 Views
About Robust Image Synthesis with Adaptive N D Tensor Voting
PowerPoint presentation about 'Robust Image Synthesis with Adaptive N D Tensor Voting'. This presentation describes the topic on This paper presents a method for repairing severely damaged images of natural scenes by using an adaptive N D Tensor Voting algorithm. The approach handles the mixture of textures and colors, inhomogeneity of patterns, and irregular object shapes commonly found in real-world images.. The key topics included in this slideshow are image repairing, robust image synthesis, tensor voting, computer vision, pattern recognition,. Download this presentation absolutely free.
1. Image Repairing: Robust Image Synthesis by Adaptive N D Tensor Voting IEEE Computer Society Conference on Computer Vision and Pattern Recognition Jiaya Jia, Chi-Keung Tang Jiaya Jia, Chi-Keung Tang Computer Science Department Computer Science Department The Hong Kong University of The Hong Kong University of Science and Technology Science and Technology
2. Motivation Main difficulties to repair a severely damaged image of natural scene Mixture of texture and colors Inhomogeneity of patterns Regular object shapes
3. Motivation Given as few as one image without additional knowledge, we address: How much color and shape information in the existing part is needed to seamlessly fill the hole? How good can we achieve in order to reduce possible visual artifact when the information available is not sufficient. Robust Tensor Voting method is adopted
4. Tensor Voting Review Tensors: compact representation of information Tensor encoding: 3D tensor Ball tensor: uncertainty in all directions Plate tensor: certainty of directions in a plate Stick tensor: certainty along two opposite directions
5. Tensor Voting Review Voting process is to propagate local information P Osculating circle
6. Image repairing system Input Damaged Image Texture-based Segmentation Statistical Region Merging Curve Connection Adaptive Scale Selection N N D Tensor Voting Output Repaired Image Complete Segmentation Image synthesis
7. Segmentation Segmentation JSEG [Deng and Manjunath 2001] color quantization spatial segmentation Mean shift [Comanicu and Meer 2002] Deterministic Annealing Framework [Hofmann et al 1998]
8. Texture-based Segmentation Texture-based Segmentation
9. Statistical Region Merge ( M + 1)D intensity vector for each region P i , where M is the maximum color depth in the whole image. histogram gradient if
10. Why Region Merge? Decrease the complexity of region topology Relate separate regions P 1 P 5 P 3 P 4 Damaged area P 2
11. Curve Connection 2D tensor voting method P 1 P 5 P 3 P 4 P 2 Z X P 2 P 4
12. Why Tensor Voting? The parameter of the voting field can be used to control the smoothness of the resulting curve. Adaptive to various hole shapes Small Scale Small Scale Large Scale Large Scale Without hole constraint Without hole constraint With hole constraint With hole constraint
13. P 4 Connection Sequence Topology of surrounding area of the hole can be very complex Greedy algorithm Always connect the most similar regions P 1 P 5 P 3 Damaged area P 2 P 2 and P 4 P 3 and P 5 P 1
14. Complete Segmentation
15. Image repairing system Input Damaged Image Texture-based Segmentation Statistical Region Merging Curve Connection Adaptive Scale Selection N N D Tensor Voting Output Repaired Image Complete Segmentation Image synthesis
16. N D Tensor Voting Tensor encoding Each pixel is encoded as a N D stick tensor 5 5 Stick tensor Scale N=26
17. N D Tensor Voting Voting process in N D space An osculating circle becomes an osculating hypersphere. N D stick voting field is uniform sampling of normal directions in the N D space. sample sample
18. Adaptive Scaling texture inhomogeneity in images gives difficulty to assign only one global scale N [Lindeberg et al 1996]. For each pixel i in images, we calculate: trace ( M ) measures the average strength of the square of the gradient magnitude in the window of size N i
19. Adaptive Scaling For each sample seed: Increase its scale Ni from the lower bound to the upper bound If trace ( ) < trace ( ) - where is a threshold to avoid small perturbation or noise interference, set Ni - 1 Ni and return Otherwise, continue the loop until maxima or upper bound is reached
25. Limitations Lack of samples. Meaningful and semi- regular objects.
26. Conclusion An automatic image repairing system. Region partition and merging. Curve connection by 2D tensor voting. N D tensor voting based image synthesis. Adaptive scale.