Image segmentation by sparse representation

This paper presents a fast and efficient algorithm, named sparse representation, for solving image segmentation. Given the human motion video each frame, we can regard each frame as an image. And with the theory of sparse representation, it is reasonable to assume that the background image can be represented by a low-rank, while the foreground human motion part is spatially localized, and therefore sparse. The human motion part is obtained by recovering the low-rank and sparse matrix. This process is formulated as a convex optimization problem that minimizes a weighted combination of two ℓ₁-norms, which can be efficiently solved by Augmented Lagrange Multiplier (ALM) method. The experimental results based on the HumanEva human motion dataset show that image segmentation by the proposed method achieves better results, and is more robust to noise, yet the process being much faster and efficient.