Presenter(s)

Evan W. Krieger, Saibabu Arigela

Download

Download Project (1.3 MB)

Description

Tracking objects, such as vehicles and humans, in wide area motion imagery (WAMI) is a challenging problem because of the limited pixel area and the low contrast/visibility of the target objects. We propose an approach to make automatic tracking algorithms more effective by incorporating image enhancement and super resolution as preprocessing algorithms. The enhancement process includes the stages of dynamic range compression and contrast enhancement. Dynamic range compression is performed by a neighborhood based nonlinear intensity transformation process, which utilizes a locally tuned inverse sine nonlinear function to generate various nonlinear curves based on pixel’s neighborhood information. These nonlinear curves are used to select the new intensity value for each pixel. A contrast enhancement technique is used to maintain or improve the contrast of the original image. Local contrast enhancement using surrounding pixel information aids in extracting higher number of features a detector can find in the image, and therefore, improves the automatic object detection capabilities. Secondly, the super resolution technique is performed on an area surrounding the object of interest to increase the size of the object in terms of pixels. The single image super resolution process is performed in the Fourier phase space which preserves the local structure of each pixel in order to estimate the interpolated pixels in the high resolution image. As a result, super resolution increases the sharpness of edges and allows for addition tracking features to be extracted. The combination of these two techniques provides the necessary preprocessing enhancement to increase the effectiveness of tracking algorithms. A quantitative evaluation is performed to compare the results of the tracking with and without the proposed techniques. The analysis is based on results of an automatic detection and tracking technique, Gaussian Ringlet Intensity Distribution (GRID), evaluated using wide area motion imagery data.

Publication Date

4-9-2014

Project Designation

Graduate Research

Primary Advisor

Vijayan Asari

Primary Advisor's Department

Vision Lab

Keywords

Stander Symposium project

Disciplines

Arts and Humanities | Business | Education | Engineering | Life Sciences | Medicine and Health Sciences | Physical Sciences and Mathematics | Social and Behavioral Sciences

Nonlinear Image Enhancement and Super Resolution for Enhanced Object Tracking

Share

COinS