PROJECT TITLE :

Robust 2D Principal Component Analysis: A Structured Sparsity Regularized Approach

ABSTRACT:

Principal element analysis (PCA) is widely used to extract features and cut back dimensionality in various laptop vision and image/video processing tasks. Standard approaches either lack robustness to outliers and corrupted information or are designed for one-dimensional signals. To handle this drawback, we propose a sturdy PCA model for 2-dimensional pictures incorporating structured sparse priors, referred to as structured sparse second-PCA. This robust model considers the previous of structured and grouped pixel values in two dimensions. As the proposed formulation is jointly nonconvex and nonsmooth, that is tough to tackle by joint optimization, we have a tendency to develop a 2-stage alternating minimization approach to resolve the matter. This approach iteratively learns the projection matrices by bidirectional decomposition and utilizes the proximal method to get the structured sparse outliers. By considering the structured sparsity prior, the proposed model becomes less sensitive to noisy knowledge and outliers in two dimensions. Moreover, the computational cost indicates that the strong 2-dimensional model is capable of processing quarter common intermediate format video in real time, also handling large-size pictures and videos, which is often intractable with alternative strong PCA approaches that involve image-to-vector conversion. Experimental results on robust face reconstruction, video background subtraction data set, and real-world videos show the effectiveness of the proposed model compared with typical 2d-PCA and alternative strong PCA algorithms.