Super-resolution of images using compressive sensing

• Main Authors: Bachini, Lawrence Ralph O., Dellomos, Danielle C., Lorilla, Marc Russell V.
• Format: text
• Language: English
• Published: Animo Repository 2014
• Subjects: High resolution imaging; Compressed sensing (Telecommunication); Signal processing > Digital techniques; Computer Engineering
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/11353
• Institution: De La Salle University

Signal reconstruction has been long tackled by researchers several decades past even up until this very moment. This has been no doubt a topic of interest by many. Ideally, for a successful signal recovery, the original signal must have no frequencies above one-half the sampling frequency, as stated by the Nyquist-Shannon sampling theory. However, this has been proven untrue by some researchers as they have discussed that a signal can still be recovered with fewer samples than the sampling theorem requires. This they called the compressive sensing. In recent years, compressive sensing has been used in super-resolution where it aims to reconstruct a low resolution image to obtain its high resolution version with a few liner combinations of basis signals. This research study aims to develop a novel algorithm to perform the same idea. Our proposed algorithm include dictionary learning using a modified K-SVD algorithm and sparse coding technique using LASCO. The novel technique in our algorithm is the feature extraction using least squares filter used to extract image information. Our method will be evaluated using quality and performance metrics and will be compared to the state-of-the-art methods. Results revealed that even though our method did not outperform the state-of-the-art, except for speed, numerical results obtained by our method are very close with the other algorithms. This implies that our method can stand on par with the state-of-the-art.