Nonconvex Energy Minimization with Unsupervised Line Process Classifier for Efficient Piecewise Constant Signals Reconstruction

In this paper, we focus on the problem of signal smoothing and step-detection for piecewise constant signals. This is central to several applications such as human activity analysis, speech or image anomaly detection in genetics. We present a two-stage approach minimize well-known line process model which arises from probabilistic representation its segmentation. first stage, TV least square detect majority continuous edges. second apply combinatorial algorithm filter all false jumps introduced by solution. The performances proposed method were tested synthetic examples. comparison recent step-preserving denoising algorithms, acceleration presents superior speed competitive quality.