Alpha channel estimation in high resolution images and image sequences

Abstract

For Motion Picture Special Effects, it is often necessary to take a source image of an acto& segment the actor from the unwanted background, and then composite over a new background. The standard approach requires the unwanted background to be a blue screen. While this technique is capable of handling areas where the foreground blends into the background, the physical requirements present many practical problems. This paper presents an algorithm that requires minimal human interaction to segment Motion Picture resolution images and image sequences. We show that it can be used not only to segment badly lit or noisy bluescreen images, but also to segment actors where the background is more varied.

Video matting of complex scenes

“…Given these distributions, the Bayesian matting approach solves for the maximum-likelihood foreground, background, and alpha at the unknown pixel. Hillman et al [2001] have additionally applied their alpha matting approach to moving image sequences. They match a low-resolution version of the current image to the previous image and classify each pixel as foreground or background if the corresponding pixels are mostly of the same foreground/background class.…”

Non-Parametric patch based video matting

“…We have used Hillman [9], Poisson [15], Closed-form [10] and Robust matting [17] algorithms for comparison. Our matting technique is applied in two different ways for analysis:…”

“…These distributions are then used to estimate all the matting variables. Hillman et al [9], improving on the idea of Ruzon and Tomasi, modeled the known local pixels as anisotropic Gaussian clusters. They used principal components analysis technique to find the major axis of these cluster which are then used for pixel-wise estimation of foreground, background colour and alpha value.…”

Temporally coherent video matting

“…But, this approach requires a specially equipped studio. Recently several interactive image matting techniques [3,4,5,6,7] have been introduced to avoid chromakeying. They allow a user to identify the background, foreground, and unknown regions of an image, in an interactive procedure which creates a segmentation called a trimap.…”