Pyramid Matting: A Resource-Adaptive Multi-Scale Pixel Pair Optimization Framework for Image Matting

Liang, Yihui; Feng, Fujian; Cai, Zixing

doi:10.1109/access.2020.2995207

Cited by 13 publications

(9 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In this experiment, a comparative analysis was conducted between GP-MCMatting and the state-of-the-art evolutionary optimization-based algorithms, such as the pyramid matting framework (PMF) [ 25 ], adaptive convergence speed controller based on particle swarm optimization (PSOACSC) [ 33 ], and the multi-objective evolutionary algorithm based on multi-criteria decomposition (MOEAMCD) [ 24 ], based on 1%, 2%, 5%, 10%, 20%, and 100% computing resources. The matting performance of the proposed algorithm under limited computing resources was verified.…”

Section: Experiments and Resultsmentioning

confidence: 99%

“…The objective function of the natural image matting problem is typically complex due to the simultaneous consideration of the similarity between pixels in the foreground and background regions, as well as the similarity among pixels. Traditional objective functions usually require nonlinear optimization algorithms for solving, which demands intensive computational resources and time [ 25 ]. When computing resources are limited, it may be challenging to directly use traditional objective function-based methods.…”

Section: Multi-criterion Matting Algorithm Via Gaussian Processmentioning

confidence: 99%

“…This algorithm utilizes all available computing resources by adjusting the number of iterations of the evolutionary algorithm and theoretically eliminates the risk of missing real samples. On the other hand, to address the problem of computing resource consumption, Liang et al [ 25 ] developed a multi-scale evolutionary pixel pair optimization framework, which transforms the large-scale pixel-pair optimization problem into multiple sub-optimization problems of different scales by using an image pyramid. Although evolutionary optimization-based methods utilizing various evolutionary algorithms have improved the accuracy of the alpha matte, it may require thousands of iterations to find the optimal solution.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Multi-Criterion Sampling Matting Algorithm via Gaussian Process

et al. 2023

View full text Add to dashboard Cite

Natural image matting is an essential technique for image processing that enables various applications, such as image synthesis, video editing, and target tracking. However, the existing image matting methods may fail to produce satisfactory results when computing resources are limited. Sampling-based methods can reduce the dimensionality of the decision space and, therefore, reduce computational resources by employing different sampling strategies. While these approaches reduce computational consumption, they may miss an optimal pixel pair when the number of available high-quality pixel pairs is limited. To address this shortcoming, we propose a novel multi-criterion sampling strategy that avoids missing high-quality pixel pairs by incorporating multi-range pixel pair sampling and a high-quality sample selection method. This strategy is employed to develop a multi-criterion matting algorithm via Gaussian process, which searches for the optimal pixel pair by using the Gaussian process fitting model instead of solving the original pixel pair objective function. The experimental results demonstrate that our proposed algorithm outperformed other methods, even with 1% computing resources, and achieved alpha matte results comparable to those yielded by the state-of-the-art optimization algorithms.

show abstract

Section: Experiments and Resultsmentioning

confidence: 99%

Section: Multi-criterion Matting Algorithm Via Gaussian Processmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Multi-Criterion Sampling Matting Algorithm via Gaussian Process

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The proposed method was implemented using Python, and the WeightedCT [12], ThreeLayer [13], KL [15], IF [17], and Pyramid [34] image matting algorithms were implemented using MATLAB. The learning-based matting algorithms, namely GCA [35], IndexNet [36], Ada [21], IamAlpha [22], as well as the DeeplabV3+ [37], EmaNet [25], SenFormer [38] semantic segmentation algorithms were implemented based on Pytorch.…”

Section: Datasets and Implementation Detailsmentioning

confidence: 99%

“…In this experiment, GCA [35] and IndexNet [36] were the chosen learning-based matting algorithms, along with IF Matting [17] as a propagation-based image matting algorithm, and Pyramid matting [34] as a optimisation-based matting algorithm. Table 4 shows the image matting performance comparison of different types of image matting algorithms before and after applying CTE-OC on the Composition-1k test set.…”

Section: Application Of Cte-oc In Different Types Of Image Matting Al...mentioning

confidence: 99%

Coarse trimap expansion based on one‐class classification for image matting

Liang

Zhu

Deng

et al. 2022

IET Image Processing

View full text Add to dashboard Cite

Trimap is required for most image matting algorithms, as it provides partial regions with known opacity. As capturing elaborate trimaps is a time‐consuming process, in practice, users prefer to provide coarse trimaps with large unknown regions. However, extant image matting algorithms cannot provide high‐quality alpha mattes based on coarse trimaps. Although some matting algorithms include trimap expansion in the pre‐processing stage, if this is done by directly comparing the similarity of image features between pixels, errors and omissions can easily occur. To overcome this issue, in this paper, a coarse trimap expansion model based on one‐class classification is presented, in which the problem is treated as a process of reclassifying pixels in unknown regions. For this purpose, a coarse trimap expansion method denoted as CTE‐OC is proposed, in which the similarity between pixels is reliably determined by measuring semantic features, allowing newly developed one‐class classifiers to adequately classify pixels in entire unknown regions. The validity of these strategies is tested experimentally, and the results show that CTE‐OC can significantly improve the quality of alpha mattes obtained by extant image matting methods when provided with coarse trimaps.

show abstract