Ramirez, Gabriel E.

Profile Picture

Filters

20141
Reset filters

Settings

Citations

Publication Search Results

Now showing 1 - 1 of 1
  • Thumbnail Image
    Publication
    Design of a compressed sensing system for hyper spectral imaging
    (2014) Ramirez, Gabriel E.; Manian, Vidya; College of Engineering; Jiménez Rodríguez, Luis O.; Morales Tirado, Lizdabel; Department of Electrical and Computer Engineering; Lorenzo Gonzalez, Edgardo
    The advent of single pixel imaging brings the promise of lower sensor costs and higher efficiency by implementing spatial compression at the same time the image is sensed. This is achieved with the use compressed sensing (CS) principles and digital micro-mirror devices. Remote sensing systems with hyper spectral imaging capabilities could benefit greatly from sensor cost reduction and more computationally efficient compression techniques. However in order to analyze the image captured with a single pixel camera it must first be reconstructed, which is a computationally intensive process. Typical remote sensing applications like surveillance or target detection where a large number of images have to be analyzed, most of which will show to be of no great interest after the fact, are ill suited for these imaging systems as most of the time would be spent reconstructing images that will later prove to be of low significance. In an effort to bring the benefits promised by single pixel imaging closer to these applications the present work develops the theory, design and implementation of CS on hyper spectral imaging. Progressive implementations of CS are performed on images, starting with spectral compression, followed by spatial compression and culminating with the proposal of a spatial CS implementation that allows recursive 2 stage reconstructions. The proposed system can be implemented on single pixel cameras while reducing the amount of computing power and time required by plain CS implementations to perform image reconstruction. The introduction of a partially recovered image also allows for a preliminary analysis of the image, allowing applications to determine if the image needs any further analysis before engaging in full image reconstruction. The partially reconstructed image is an arrangement of the measurements from spatially compressed image sections. The final compression stage can tackle each section as a separate image to be reconstructed; this is achieved using a structured measurement matrix. The following chapters provide tests and experiments that compare processing times, classification statistics and error rates that point towards the systems practicality, making it an interesting option for signals with high data volumes like those found in hyper spectral imaging.