Fast Multi-Level Foreground Estimation

Germer, Thomas A.; Uelwer, Tobias; Conrad, Stefan; Harmeling, Stefan

doi:10.1109/icpr48806.2021.9412408

Cited by 4 publications

(3 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…2. For the variants (FG w/ [13]) and (FG w/ [12]), they report similar results to the complete model (w * , n * ), which indicates that our method is not sensitive to the foreground estimation methods. The variant (FG w/ GT α) presents a similar performance with foreground estimation using the enhanced matte (w * , n * ).…”

Section: F Ablation Studymentioning

confidence: 53%

“…We first compare with six variants: 1) optimizing the latent code w only while keeping the noise code n fixed (w * , n); 2) optimizing the noise code n only (w, n * ); 3) without the entropy #2). Furthermore, for evaluating the influence of different foreground estimation methods, we use MLFgE [13] and Blur-Fusion [12] instead of Closed-form Matting [27] for the foreground estimation respectively, resulting in two variants: 9) FG w/ [13] and 10) FG w/ [12]. 11) We also use the GT matte instead of enhanced matte for the foreground estimation used in Eq.…”

Section: F Ablation Studymentioning

confidence: 99%

See 1 more Smart Citation

Self-Supervised Matting-Specific Portrait Enhancement and Generation

Zhou

2022

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

We resolve the ill-posed alpha matting problem from a completely different perspective. Given an input portrait image, instead of estimating the corresponding alpha matte, we focus on the other end, to subtly enhance this input so that the alpha matte can be easily estimated by any existing matting models. This is accomplished by exploring the latent space of GAN models. It is demonstrated that interpretable directions can be found in the latent space and they correspond to semantic image transformations. We further explore this property in alpha matting. Particularly, we invert an input portrait into the latent code of StyleGAN, and our aim is to discover whether there is an enhanced version in the latent space which is more compatible with a reference matting model. We optimize multi-scale latent vectors in the latent spaces under four tailored losses, ensuring matting-specificity and subtle modifications on the portrait. We demonstrate that the proposed method can refine real portrait images for arbitrary matting models, boosting the performance of automatic alpha matting by a large margin. In addition, we leverage the generative property of StyleGAN, and propose to generate enhanced portrait data which can be treated as the pseudo GT. It addresses the problem of expensive alpha matte annotation, further augmenting the matting performance of existing models. Code is available at https://github.com/cnnlstm/StyleGAN Matting.

show abstract

Section: F Ablation Studymentioning

confidence: 53%

Section: F Ablation Studymentioning

confidence: 99%

Self-Supervised Matting-Specific Portrait Enhancement and Generation

Zhou

2022

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

show abstract

“…Our method achieves more accurate matting results, especially around thin and detailed structures, compared to prior work. We extracted the foreground using the technique proposed by Germer et al[Germer et al 2021] and composited it onto a new background sampled from a public background database].…”

mentioning

confidence: 99%

Matting by Generation

Wang,

Li,

Wang

et al. 2024

Special Interest Group on Computer Graphics and Interactive Techniques Conference Conference Papers '24

View full text Add to dashboard Cite

diffusion models, enriched with extensive pre-trained knowledge, to regularize the matting process. We present novel architectural innovations that empower our model to produce mattes with superior resolution and detail. The proposed method is versatile and can perform both guidance-free and guidance-based image matting, accommodating a variety of additional cues. Our comprehensive evaluation across three benchmark datasets demonstrates the superior performance of our approach, both quantitatively and qualitatively. The results not only reflect our method's robust effectiveness but also highlight its ability to generate visually compelling mattes that approach photorealistic quality. The project page for this paper is at https://lightchaserx.github.io/matting-by-generation/.

show abstract

Approximate Fast Foreground Colour Estimation

Forte¹

2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Compositing objects from one image or video onto another background is a classic computer vision problem. Compositing requires an estimate of the foreground object's transparency, a topic of study known as Alpha Matting. However, less studied is the requirement of an estimate for the foreground object's colours in transparent regions. Naively compositing without this leads to bleed-through of the original background. In this work, we propose an approximation for the multi-level foreground estimation technique from Germer et al. [1]. Our method can be implemented 11 lines of Python code and yields comparable results to state of the art approaches while also being faster. Our code is available at https://github.com/PhotoRoom/ fast-foreground-estimation.

show abstract

Fast Multi-Level Foreground Estimation

Cited by 4 publications

References 16 publications

Self-Supervised Matting-Specific Portrait Enhancement and Generation

Self-Supervised Matting-Specific Portrait Enhancement and Generation

Matting by Generation

Approximate Fast Foreground Colour Estimation

Contact Info

Product

Resources

About