Despeckling of SAR Images Using Residual Twin CNN and Multi-Resolution Attention Mechanism

Blaž Pongrac (Author), Dušan Gleich (Author)

Abstract

The despeckling of synthetic aperture radar images using two different convolutional neural network architectures is presented in this paper. The first method presents a novel Siamese convolutional neural network with a dilated convolutional network in each branch. Recently, attention mechanisms have been introduced to convolutional networks to better model and recognize features. Therefore, we propose a novel design for a convolutional neural network using an attention mechanism for an encoder–decoder-type network. The framework consists of a multiscale spatial attention network to improve the modeling of semantic information at different spatial levels and an additional attention mechanism to optimize feature propagation. Both proposed methods are different in design but they provide comparable despeckling results in subjective and objective measurements in terms of correlated speckle noise. The experimental results are evaluated on both synthetically generated speckled images and real SAR images. The methods proposed in this paper are able to despeckle SAR images and preserve SAR features.

Keywords

radarji;globoko učenje;konvolucijske nevronske mreže;synthetic aperture radar;speckle;speckle suppression;despeckling;deep learning;convolutional neural network;

Data

Language:	English
Year of publishing:	2023
Typology:	1.01 - Original Scientific Article
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	MDPI
UDC:	621.39
COBISS:	161242883
ISSN:	2072-4292
Views:	280
Downloads:	17
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	Slovenian
Secondary abstract:	V tem članku smo predstavilo odstranjevanje pegastega šuma slik s sintetično odprtino z uporabo dveh različnih arhitektur konvolucijskih nevronskih mrež. Prva metoda predstavlja novo siamsko konvolucijsko nevronsko mrežo z razširjeno konvolucijsko mrežo v vsaki veji. Nedavno so bili v konvolucijska omrežja uvedeni mehanizmi pozornosti za boljše modeliranje in prepoznavanje funkcij. Zato predlagamo novo zasnovo za konvolucijsko nevronsko mrežo z uporabo mehanizma pozornosti za omrežje tipa kodirnik dekodirnik. Ogrodje je sestavljeno iz večstopenjskega omrežja prostorske pozornosti za izboljšanje modeliranja semantičnih informacij na različnih prostorskih ravneh in dodatnega mehanizma pozornosti za optimizacijo širjenja značilnosti. Obe predlagani metodi se razlikujeta po zasnovi, vendar zagotavljata primerljive rezultate odstranjevanja pegastega šuma pri subjektivnih in objektivnih meritvah v smislu koreliranega pegastega šuma. Eksperimentalni rezultati so ovrednoteni tako na sintetično ustvarjenih pegastih slikah kot na resničnih slikah SAR. Metode, predlagane v tem dokumentu, lahko odstranijo pegasti šum iz slik SAR in ohranijo bistvene lastnosti slik SAR.
Secondary keywords:	radarji;globoko učenje;konvolucijske nevronske mreže;
Type (COBISS):	Scientific work
Pages:	25 str.
Volume:	ǂVol. ǂ15
Issue:	ǂiss. ǂ14, [article no.] 3698
Chronology:	2023
DOI:	10.3390/rs15143698
ID:	22967190