This thesis discusses problems related to imaging applications, focusing on image compression, transmission and source camera identification. Random projections and compressed sensing are shown to be effective techniques to address such problems. We show that random projections can be used to compress multispectral images in a more rate-efficient manner than previously known and with a low-complexity scheme based on compressed sensing that is suitable for usage onboard of spacecrafts. Image measurements obtained through random projections possess peculiar properties that allow to devise novel schemes to mitigate channel unreliability when the data have to be transmitted over such channels. This is shown to be particularly useful for signal acquisition and transmission on low-power devices. Finally, quantized random projections implement low dimensionality signal embeddings that, by approximately preserving the geometry of the original signal space, allow to process high volumes of information with reduced complexity. We show that they can be used to address the problem of compressing camera sensor fingerprints, significantly improving over state of the art methods in terms of storage requirements and matching speed. Such improvements enable the development of novel applications such as camera-based image retrieval on large scales, for which we study efficient search algorithms exploiting properties of random projections.

Imaging using random projections: compression, communication, camera identification / Valsesia, Diego. - (2016).

Imaging using random projections: compression, communication, camera identification

VALSESIA, DIEGO
2016

Abstract

This thesis discusses problems related to imaging applications, focusing on image compression, transmission and source camera identification. Random projections and compressed sensing are shown to be effective techniques to address such problems. We show that random projections can be used to compress multispectral images in a more rate-efficient manner than previously known and with a low-complexity scheme based on compressed sensing that is suitable for usage onboard of spacecrafts. Image measurements obtained through random projections possess peculiar properties that allow to devise novel schemes to mitigate channel unreliability when the data have to be transmitted over such channels. This is shown to be particularly useful for signal acquisition and transmission on low-power devices. Finally, quantized random projections implement low dimensionality signal embeddings that, by approximately preserving the geometry of the original signal space, allow to process high volumes of information with reduced complexity. We show that they can be used to address the problem of compressing camera sensor fingerprints, significantly improving over state of the art methods in terms of storage requirements and matching speed. Such improvements enable the development of novel applications such as camera-based image retrieval on large scales, for which we study efficient search algorithms exploiting properties of random projections.
2016
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2642257
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo