Non-parametric natural image matting

Sarim, Muhammad; Hilton, Adrian; Guillemaut, Jean-Yves; Kim, Han Sung

doi:10.1109/icip.2009.5414367

Cited by 5 publications

(10 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Supposed e contains two evidences E1 and E2, the corresponding Basic Probability Assignment are ml and m2, whose focal elements are A, and Bj respectively. If K < 1 , the formula of D-S evidence theory combination is established as (3) where K = I � (AJm2 (B) , which reflects the A,nB j =<t> degree of evidences' conflict [16]. In fact, many evidences to the contrary have existed.…”

Section: Combination Rule Of D-s Evidence Theorymentioning

confidence: 99%

Matting image based on D-S evidence theory

Lei

Zuo

Cao

2013

IEEE Conference Anthology

View full text Add to dashboard Cite

This paper presents a new method for the matting problem of color image, which is based on the ideology of image matting proposed by Thomas Porter and Tom Duff and the algorithm of D-S evidence theory. Firstly, the given image must be marked as many approaches. The difference from other methods is that the boundary of foreground called uncertain region should only be signed as constraints in this paper. Then 10 features about the candidate pixel, which is one pixel of the uncertain region, are extracted from the corresponding color space as the inputs of D-S evidence theory and partitioned into two conflictive subsets. Moreover, the results of the two subsets computed by the combination rule of D-S evidence theory should be synthesized again to get S according to their confidence levels. Then the comparativeresult of s and (5 (0.5) is used to determine whether the candidate pixel belongs to foreground or not. If S > (5 , the candidate pixels considered as a foreground's pixel. On the contrary, it's a background's pixel. Finally, some images are downloaded from internet and matted by using this paper's method, and the results show that this method performs well.

show abstract

Section: Combination Rule Of D-s Evidence Theorymentioning

confidence: 99%

Matting image based on D-S evidence theory

Lei

Zuo

Cao

2013

IEEE Conference Anthology

View full text Add to dashboard Cite

show abstract

“…Although affinity based approaches overcome the limitation of sample misclassification they are prone to accumulation of small errors in the final alpha matte because of their propagation behaviour. Non-parametric matching technique [16] have also recently been demonstrated to achieve comparable natural image matting results allowing representation of local appearance structure which avoids smoothness assumptions.…”

Section: Related Workmentioning

confidence: 99%

“…The proposed algorithm has used a template clustering approach to propagate the trimap across views. Once a trimap is estimated an image matting algorithm [16] is used to extract the final alpha matte. The approach is limited to a single frame and neighbouring views only and does not perform well for views having a baseline > 50…”

Section: Related Workmentioning

confidence: 99%

“…The problem is then to estimate the α value for the unknown pixels given the trimap labels. A number of solutions [4,8,12,14,16,18,20] have been proposed giving high-quality alpha matte estimation. Video matting extends this approach to sequences by temporal propagation of the manually defined trimap labels for key-frames across the sequence using cues such as optical flow and foreground/background appearance [1,3].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Multiple View Wide-Baseline Trimap Propagation for Natural Video Matting

Sarim

Hilton

Guillemaut

et al. 2010

2010 Conference on Visual Media Production

View full text Add to dashboard Cite

This paper presents a method to estimate alpha mattes for video sequences of the same foreground scene from wide-baseline views given sparse key-frame trimaps in a single view. A statistical inference framework is introduced for spatio-temporal propagation of high-confidence trimap labels between video sequences without a requirement for correspondence or camera calibration and motion estimation. Multiple view trimap propagation integrates appearance information between views and over time to achieve robust labelling in the presence of shadows, changes in appearance with view point and overlap between foreground and background appearance. Results demonstrate that trimaps are sufficiently accurate to allow high-quality video matting using existing single view natural image matting algorithms.Quantitative evaluation against ground-truth demonstrates that the approach achieves accurate matte estimation for camera views separated by up to 180• , with the same amount of manual interaction required for conventional single view video matting.

show abstract

“…The problem is then to estimate the α value for the pixels in the unknown region given the definite trimap labels. Several solutions [2,3,4,5] have been proposed to estimate a high quality alpha matte by exploiting the statistics of the definite regions. In recent techniques like [6,7,8,9,10], single view matting approaches have been extended to narrow-baseline multiple views by assuming primarily the invariant inter-view foreground appearance.…”

Section: Introductionmentioning

confidence: 99%

Natural image matting for multiple wide-baseline views

Sarim

Hilton

Guillemaut

et al. 2010

2010 IEEE International Conference on Image Processing

Self Cite

View full text Add to dashboard Cite

In this paper we present a novel approach to estimate the\ud alpha mattes of a foreground object captured by a widebaseline\ud circular camera rig provided a single key frame\ud trimap. Bayesian inference coupled with camera calibration\ud information are used to propagate high confidence trimaps\ud labels across the views. Recent techniques have been developed\ud to estimate an alpha matte of an image using multiple\ud views but they are limited to narrow baseline views with low\ud foreground variation. The proposed wide-baseline trimap\ud propagation is robust to inter-view foreground appearance\ud changes, shadows and similarity in foreground/background\ud appearance for cameras with opposing views enabling high\ud quality alpha matte extraction using any state-of-the-art image\ud matting algorithm

show abstract

Non-parametric natural image matting

Cited by 5 publications

References 10 publications

Matting image based on D-S evidence theory

Matting image based on D-S evidence theory

Multiple View Wide-Baseline Trimap Propagation for Natural Video Matting

Natural image matting for multiple wide-baseline views

Contact Info

Product

Resources

About