Object Detection in Degraded Visual Environments using Compressive Sensing

摘要

Compressive Sensing (CS) has proven its ability to reduce the number of measurements required to reproduce images with similar quality to those reconstructed by observing the Shannon-Nyquist sampling criteria. By exploiting spatial redundancies, it was shown that CS can be used in target recovery and object detection. In this paper we propose a method that incorporates an effcient use of CS to locate a specific object in zero-visibility environments. We show that with the use of an over-complete dictionary of the target our technique can perceive the location of the target from hidden information in the scene. This paper will compare previously implemented algorithms with our, list the shortcomings evident in their outputs, explain our setups, detail the differences in dictionary structures, and present quantified results to support its efficacy in the results section.