Image foreground matting method based on neighborhood and non-neighborhood smoothness priors

摘要

The present invention discloses an image foreground matting method based on neighborhood and non-neighborhood smoothness priors. The method primarily comprises the steps of: interactively marking foreground points and background points; initializing α values of each unmarked pixel of the input image by a color sampling method, calculating confidence degree of the pixel, and admitting α values of pixels of which confidence degree is larger than a given threshold as known pixels; calculating data term weights, neighborhood smoothness constraint term weights and non-neighborhood smoothness constraint term weights of each pixel in the input image to construct graph patterns of all pixels of the input image; and according to α values of the known pixels, under the constraint of the graph patterns, solving probabilities that each pixel belongs to the foreground by minimizing the energy equation so as to obtain alpha mattes.

主权项

1. An image foreground matting method based on neighborhood and non-neighborhood smoothness priors, comprising the steps of:
step S100, marking a foreground area, a background area, and an unknown area in the input image; step S200, initializing probabilities α values that a pixel belongs to the foreground by a color sampling method for each pixel in the unknown area of the input image, calculating confidence degree of α values, admitting α values of pixels of which confidence degree is larger than a given threshold, marking pixels as known pixels, setting α values of each pixel in the foreground area to a maximum value, and setting α values of each pixel in the background area to a minimum value; step S300, calculating data term weights of each pixel in the input image according to α values of each pixel, calculating neighborhood smoothness constraint term weights and non-neighborhood smoothness constraint term weights of each pixel, and constructing an overall graph patterns for all pixels of the input image according to three kinds of weights; step S400, according to α values of all foreground area pixels, background area pixels and the known pixels in the unknown area, under a constraint of graph patterns from step S300, solving probabilities that each pixel belongs to the foreground by minimizing an energy equation so as to obtain alpha mattes.