The Weight-Shape decomposition of density estimates: A framework for clustering and image analysis algorithms

作者：

Highlights：

• Every Parzen estimate of a probability distribution can be decomposed into Weight and Shape components in a unique fashion.

• This decomposition allows for a fresh classification of clustering algorithms, based on Probability (mean-shift), on Shape (quantum clustering), or on Weight (entropy maximization).

• These concepts can be applied to image analysis, with Shape providing novel aspects of generalized edge detection.

摘要

•Every Parzen estimate of a probability distribution can be decomposed into Weight and Shape components in a unique fashion.•This decomposition allows for a fresh classification of clustering algorithms, based on Probability (mean-shift), on Shape (quantum clustering), or on Weight (entropy maximization).•These concepts can be applied to image analysis, with Shape providing novel aspects of generalized edge detection.

论文关键词：Density estimate,Quantum clustering,Mean-shift clustering,Maximum entropy,Image contour extraction

论文评审过程：Received 30 August 2017, Revised 20 March 2018, Accepted 27 March 2018, Available online 29 March 2018, Version of Record 18 April 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.03.034