646 IEEE JOURNAL OF OCEANIC ENGINEERING, VOL. 36, NO. 4, OCTOBER 2011 Peer-Reviewed Technical Communication FPGA-Based Adaptive Speckle Suppression Filter for Underwater Imaging Sonar Serge Karabchevsky, David Kahana, Ortal Ben-Harush, and Hugo Guterman Abstract—Underwater forward-looking imaging sonar (FLS) is widely used on stationary and moving platforms to overcome un- derwater visibility problems. The SONAR images are perturbed by a multiplicative noise called speckle, due to the coherent nature of the scattering phenomenon. Speckle reduction filters are neces- sary to optimize the images’ exploitation procedures. The results of speckle filters may vary from one sensor and one wavelength to another; therefore, no generic de-speckling algorithm exists. Sev- eral studies have been carried out on speckle noise suppression on sidescan sonar, but the problem of speckle noise suppression for FLS has not yet been covered. A comparison of the most used clas- sical speckle suppression filters as well as advanced wavelet-based ones was carried out. The Frost filter was found to be the most ade- quate for FLS data, but also the most computationally complex and not suitable for real-time processing. Two novel architectures for real-time and low-power field-programmable gate array (FPGA) implementation of the Frost speckle filter for underwater imaging sonar are presented. The proposed architectures have superior per- formance and power efficiency compared to standard software im- plementation. Index Terms—Adaptive filters, field-programmable gate array (FPGA), speckle, sonar. I. I NTRODUCTION F ORWARD-LOOKING IMAGING SONAR (FLS) is used in autonomous underwater vehicles (AUVs) and handheld platforms for obstacle avoidance, detection, and mapping [1]. These devices employ a set of acoustic transducers to produce acoustic beams at a fixed frequency. The returned beams are re- ceived by the transducers and the time difference is measured, which allows creation of a 2-D video from a 1-D transducer array [2]. When an AUV is operating in an unknown environ- ment, the success of the mission is strongly dependent on the re- liability of the obstacle detection system, which is a function of the FLS performance. Due to the coherent nature of the acoustic beams, the data produced by the sonar systems are affected by Manuscript received September 12, 2010; revised January 09, 2011 and May 03, 2011; accepted May 17, 2011. Date of publication August 12, 2011; date of current version October 21, 2011. Associate Editor: R. Eustice. S. Karabchevsky is with the Department of Electro-Optical Engineering, Ben- Gurion University of The Negev, Beer-Sheva 17906, Israel (e-mail: serge@ee. bgu.ac.il). D. Kahana,O. Ben-Harush, and H. Guterman arewith theDepart- mentof Electrical and Computer Engineering, Ben-Gurion University of the Negev,Beer-Sheva17906,Israel(e-mail:david.kahana@gmail.com; talibh201@gmail.com; hugo@ee.bgu.ac.il). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JOE.2011.2157729 speckle noise [3], decreasing the ability to detect and iden objects and obstacles in the acoustic images. AUVs can be equipped with up to two sonar devices ope at a combined frame rate of 20 fps, while in the handheld forms a single sonar is used, operating at frame rate of up fps [4]. Sonar images are around 600 pixels high by 500 p wide (sonar range dependent). Serious power limitations e the handheld device is the most critical in power consump it uses a 12-V battery with 2.5-Ah capacity that provides 2 of operation for a 15-W system [5] (including a 10-W SON [4]). The power budget for AUVs is typically limited to 100 for the data processing system and every watt counts for t sion duration. While the FLS is used for obstacle detection purposes, the most commonly used sonar in underwater applications is s the sidescan sonar (SSS). SSS is not employed for obstacl tection purposes as it is mounted on the sides of the AUV. SSS uses a single beam transducer transmitting a single a beam from both sides of the AUV, which is broad in a vert dimension (40 –60 ) and narrow in the along-track dimension ( 2.5 mm) to minimize the overlap between successive pin As with the line scanner, each ping produces a single line the output image as the vehicle advances on its path, whil data across a line are generated from the relative time of backscatter return. Additional adjustments are done on th data to overcome the distortions caused by the vehicle ori tion and geometry [6]. Due to the scanning method, the lin the image produced by the SSS are independent and ident distributed in terms of speckle noise and can be treated se rately. The forward-looking sonar, on the other hand, uses an array of transducers (up to 256 [4]), which covers a sector cally 45 15 and produces a 2-D image in a single ping. Th wide angles of the sector and the number of transducers i duce wave interference not observed in the SSS. While se studies have been carried out on speckle noise suppressio SSS [7]–[9], the problem of speckle noise suppression for to the best of our knowledge, has not yet been covered. Speckle noise suppression techniques can be divided int categories. The first approach is to average several looks ages acquired of the same scene, which is called multilook cessing or super-resolution [10]. If several images are acq from the same scene and then averaged, the speckle noise reduced due to its random nature, while the observed scen not be degraded. The second technique is based on filterin speckle noise after the images are formed. The simplest fi that can be applied on the speckled image are low-pass (b 0364-9059/$26.00 © 2011 IEEE