Naslov
Retrieving Scene Depth from Defocused Image Using Compressed Sensing

Autori
Jurišić Bellotti, Maja ; Vučić, Mladen

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, prošireni sažetak, znanstveni

Izvornik
Abstract Book of Third International Workshop on Data Science / - Zagreb, 2018, 12-13

Skup
3rd International Workshop on Data Science (IWDS 2018)

Mjesto i datum
Zagreb, Hrvatska, 16.10.2018

Vrsta sudjelovanja
Poster

Vrsta recenzije
Domaća recenzija

Ključne riječi
compressed sensing ; range imaging ; sparse signal processing

Sažetak
Information about object’s distance from the camera is scrambled in the defocused image. Retrieving this information is an ill posed problem which can be solved using compressed sensing. Imaging system captures the surface of the observed object which in our case consists of the point light sources placed on a grid. This surface occupies a small part of the entire object. Therefore, the observed signal is sparse in the spatial domain. We use this sparsity to form a compressed sensing problem in which the observed signal represents sensor’s pixel values, and unknown signal represents the luminous intensities of possible object light sources in a 3D-volume. For such a system, we discuss the properties of the measurement matrix necessary for a unique reconstruction. For lowering the computational complexity we divide the image into smaller subimages and solve optimization problems separately. Using simulated experiments, we illustrate that scene depth information can be extracted from a defocused image.

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika, Računarstvo

POVEZANOST RADA