Weighted color and texture sample selection for image matting
Weighted Color and Texture Sample Selection for
Image Matting
AIM:
The mainstay of the project is a new non-parametric sampling based method is presented that uses texture as an additional feature for the matting task.
INTRODUCTION:
Digital matting refers to the accurate extraction of foreground objects from an image here part of the regions in the object could have contributions from the background. This contribution is incorporated into a compositing equation in the form of opacity α of a pixel; the equation expresses the observed color value of a pixel as a convex combination of foreground (F) and background (B) colors. The opacity takes value in the range [0, 1], with 0 indicating that the pixel is from the background and 1 indicating that it is from the foreground. Estimating the digital matte is useful in image and video editing tasks such as background replacement.
BLOCK DIAGRAM:
EXISITING SCHEME:
Poisson matting, Closed-form matting and Non local matting. The assumptions of large kernels by and local color line of are relaxed in KNN matting using non local principles and K nearest neighbors, SVR Matting, Bayesian Matting, and Robust Matting.
PROPOSED SCHEME:
In this paper, a new non-parametric sampling based method is presented that uses texture as an additional feature for the matting task. Our sampling strategy considers both local boundary as well as global samples, where the former implies samples collected from the boundaries of known regions while the latter implies that samples are collected from diverse locations in the image. Local and global candidate sets of F and B samples are collected and a new objective function is formulated that combines color and texture statistics to determine the best samples that represent the true foreground and background of unknown pixels. Finally, the estimated mattes are refined using conventional Laplacian approach.
MODULES:
MODULE I
Region Selection:
Input Image foreground and
Image Matting
AIM:
The mainstay of the project is a new non-parametric sampling based method is presented that uses texture as an additional feature for the matting task.
INTRODUCTION:
Digital matting refers to the accurate extraction of foreground objects from an image here part of the regions in the object could have contributions from the background. This contribution is incorporated into a compositing equation in the form of opacity α of a pixel; the equation expresses the observed color value of a pixel as a convex combination of foreground (F) and background (B) colors. The opacity takes value in the range [0, 1], with 0 indicating that the pixel is from the background and 1 indicating that it is from the foreground. Estimating the digital matte is useful in image and video editing tasks such as background replacement.
BLOCK DIAGRAM:
EXISITING SCHEME:
Poisson matting, Closed-form matting and Non local matting. The assumptions of large kernels by and local color line of are relaxed in KNN matting using non local principles and K nearest neighbors, SVR Matting, Bayesian Matting, and Robust Matting.
PROPOSED SCHEME:
In this paper, a new non-parametric sampling based method is presented that uses texture as an additional feature for the matting task. Our sampling strategy considers both local boundary as well as global samples, where the former implies samples collected from the boundaries of known regions while the latter implies that samples are collected from diverse locations in the image. Local and global candidate sets of F and B samples are collected and a new objective function is formulated that combines color and texture statistics to determine the best samples that represent the true foreground and background of unknown pixels. Finally, the estimated mattes are refined using conventional Laplacian approach.
MODULES:
MODULE I
Region Selection:
Input Image foreground and