Fast cost-volume filtering for visual correspondence and beyond

Rhemann, Christoph; Hosni, Asmaa; Bleyer, Michael; Rother, Carsten; Gelautz, Margrit

doi:10.1109/cvpr.2011.5995372

Cited by 453 publications

(461 citation statements)

References 16 publications

Supporting

Mentioning

448

Contrasting

Unclassified

Order By: Relevance

“…Although exact inference in such models is in general intractable, much attention has been paid to developing fast approximation algorithms, including variants of belief propagation, dual decomposition methods, and move-making approaches [1][2][3]. Recently, a number of cross bilateral Gaussian filter-based methods have been proposed for problems such as object class segmentation [4], denoising [5], stereo and optical flow [6],…”

Section: Introductionmentioning

confidence: 99%

Filter-Based Mean-Field Inference for Random Fields with Higher-Order Terms and Product Label-Spaces

Warrell²,

2014

View full text Add to dashboard Cite

Abstract. Recently, a number of cross bilateral filtering methods have been proposed for solving multi-label problems in computer vision, such as stereo, optical flow and object class segmentation that show an order of magnitude improvement in speed over previous methods. These methods have achieved good results despite using models with only unary and/or pairwise terms. However, previous work has shown the value of using models with higher-order terms e.g. to represent label consistency over large regions, or global co-occurrence relations. We show how these higher-order terms can be formulated such that filter-based inference remains possible. We demonstrate our techniques on joint stereo and object labeling problems, as well as object class segmentation, showing in addition for joint object-stereo labeling how our method provides an efficient approach to inference in product label-spaces. We show that we are able to speed up inference in these models around 10-30 times with respect to competing graph-cut/move-making methods, as well as maintaining or improving accuracy in all cases. We show results on PascalVOC-10 for object class segmentation, and Leuven for joint object-stereo labeling.

show abstract

Section: Introductionmentioning

confidence: 99%

Filter-Based Mean-Field Inference for Random Fields with Higher-Order Terms and Product Label-Spaces

Warrell²,

2014

View full text Add to dashboard Cite

show abstract

“…이 과 정에서 자연스러운 영상들을 얻기 위해 다시점 영상 의 정렬화 [3] , 색상 보정 [4] 작업이 수행된다. 의 논문에서는 영상의 색상 정보와 기울기 정보를 이 용한 화소 단위의 비용 계산 방법을 사용했다 [7] . 초기 조를 이용한 방법이 사용됐다 [6] .…”

Section: 다시점 영상을 얻기 위해 많은 수의 카메라로 영상을 촬영하기도 하지만 스테레오 영상을 이용해 가상시점unclassified

Stereo Matching Using Distance Trasnform and 1D Array Kernel

Chang¹,

Ho²

2016

The Journal of Korean Institute of Communications and Informati

View full text Add to dashboard Cite

A stereo matching method is one of the ways to obtain a depth value from two dimensional images. This method estimates the depth value of target images using stereo images which have two different viewpoints. In the result of stereo matching, the depth value is represented by a disparity value. The disparity means a distance difference between a current pixel in one side of stereo images and its corresponding point in the other side of stereo images. The stereo matching in a homogeneous region is always difficult to find corresponding points because there are no textures in that region. In this paper, we propose a novel matching equation using the distance transform to estimate accurate disparity values in the homogeneous region. The distance transform calculates pixel distances from the edge region. For this reason, pixels in the homogeneous region have specific values when we apply this transform to pixels in that region. Therefore, the stereo matching method using the distance transform improves the matching accuracy in the homogeneous regions. In addition, we also propose an adaptive matching cost computation using a kernel of one dimensional array depending on the characteristic of regions in the image. In order to aggregate the matching cost, we apply a cross-scale cost aggregation method to our proposed method. As a result, the proposed method has a lower average error rate than that of the conventional method in all regions.

show abstract

“…Furthermore, lense vignetting effect on sub-aperture images is non-negligible. Therefore, standard stereo matching methods [12] are not directly applicable as mentioned by Jeon et al [9]. Recent studies [6], [9], [10], [13] have addressed the problem caused by extremely short baselines.…”

Section: Introductionmentioning

confidence: 99%

“…The cost calculation method expressed in (6) is computationally efficient and has the robustness against sensor noise thanks to using a majority binary string for computation of the Hamming distance. After aggregating the initial matching cost over α p , we refine the result by using the cost-volume filtering [12] with the guided image filtering [18]. We select the optimum α p that minimizes the refined matching cost C(p, α p ) as an initial depth value:…”

Section: Initial Depth Estimation Using Census Transformmentioning

confidence: 99%

“…Next, a matching cost among the binarized images is computed using the majority operations between corresponding bits. In addition, we apply the cost-volume filtering [12] to the aggregated matching cost. Finally, we find the optimal value of remapped parameter α p that minimizes the matching cost to estimate the initial depth.…”

Section: Initial Depth Estimation Using Census Transformmentioning

confidence: 99%

See 1 more Smart Citation

Depth Map Estimation Using Census Transform for Light Field Cameras

Tomioka

Mishiba

Oyamada

et al. 2017

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYDepth estimation for a lense-array type light field camera is a challenging problem because of the sensor noise and the radiometric distortion which is a global brightness change among sub-aperture images caused by a vignetting effect of the micro-lenses. We propose a depth map estimation method which has robustness against sensor noise and radiometric distortion. Our method first binarizes sub-aperture images by applying the census transform. Next, the binarized images are matched by computing the majority operations between corresponding bits and summing up the Hamming distance. An initial depth obtained by matching has ambiguity caused by extremely short baselines among sub-aperture images. After an initial depth estimation process, we refine the result with following refinement steps. Our refinement steps first approximate the initial depth as a set of depth planes. Next, we optimize the result of plane fitting with an edge-preserving smoothness term. Experiments show that our method outperforms the conventional methods.

show abstract

Fast cost-volume filtering for visual correspondence and beyond

Cited by 453 publications

References 16 publications

Filter-Based Mean-Field Inference for Random Fields with Higher-Order Terms and Product Label-Spaces

Filter-Based Mean-Field Inference for Random Fields with Higher-Order Terms and Product Label-Spaces

Stereo Matching Using Distance Trasnform and 1D Array Kernel

Depth Map Estimation Using Census Transform for Light Field Cameras

Contact Info

Product

Resources

About