The Effect of Resolution on Detecting Visually Salient Preattentive Features

Report No. ARL-TR-7317
Authors: Adrienne Raglin and Christine Chan
Date/Pages: June 2015; 16 pages
Abstract: Determining the factors that affect saliency within an image is valuable to a wide variety of real-world applications, such as object detection and identification. In this report, we discuss an initial investigation on the impact of 1 factor in particular, image resolution. Understanding the relationship between image resolution and saliency may lead to new image processing techniques that are able to factor in an attempt to improve the extraction of valuable information from a scene. A total of 12 scenes, 10 natural and 2 artificial, were used to create the image set used in this study, which consisted of images of each scene at 100%, 75%, 50%, 25%, and 18.75% of the original image resolution. The Harrison and Etienne-Cummings ideal observer model (IOM) was then applied to the image set so that salient areas could be highlighted. The IOM was used to identify the 20 most salient regions within each image. Then for each scene, the performance of the IOM was compared across all resolutions. We were particularly interested in whether the same salient regions persisted across all resolutions. The results of the study show that resolution may affect the selection of salient areas within the IOM.
Distribution: Approved for public release
  Download Report ( 0.374 MBytes )
If you are visually impaired or need a physical copy of this report, please visit and contact DTIC.
 

Last Update / Reviewed: June 1, 2015