The alpha matte is a two-dimensional map that is used to combine two images, one containing a foreground and the other containing a background. Alpha matte extraction is performed on green-screen images and requires user interaction to tune parameters in different preprocessing and postprocessing stages to refine an alpha matte. This paper tackles the problem of fully automatic extraction of the foreground on green-screen images with extraction of the corresponding alpha matte. The method is based on a multi-layer perceptron that assigns an alpha value, from a discrete set of ten alpha values, to each patch on a green-screen image. The approach for assigning an alpha value to an image patch is based on a set of features that enhance discrimination between foreground and background. The classifier is trained to learn to separate foreground objects from green-screen backgrounds as well as to generate the corresponding alpha matte map required for subsequent digital compositing. To test how the proposed approach handles alpha matte extraction under unsuitable conditions, a 64-image dataset was generated. The main contribution is that our method overcomes two challenges publicly posed within a dataset of green-screen image sequences, donated by Hollywood Camera Work LLC. Tests with this dataset generate high-quality visual results for those two cases. These results are confirmed by comparing the proposed fully automatic alpha matte extraction with that based on the use of Adobe After Effects Creative Cloud, an application which heavily depends on user interaction.
展开▼
