UWIT: underwater image toolbox for optical image processing and mosaicking in MATLAB

Abstract

Absrml-This paper shows results from our development ofnn extended MATLAU image procesing toolbox, which implements some useful opticnl image processing and mosnicking algorithms found in the literature. We surveyed and selected algorithms from the field which showed promise in application to the underwnler environment. We then enended there nlgorithms to explicitly deai with the unique constraints ufunderwater imagery in the building of our toolbox. As such, the nigorithms implemented include:contrnst limited adaptive hbtogram spedficntion (CIAIIS) to den1 with the inherent nonuniform lighting in underwnter imagery.. Fourier b u d methods for sutle, rotation, und trnmlalion ncovery which provide rohustnes against dissimihr image mgiOnS.. local normalid cordstion for imnge registration to handle the low-fenlure, unstructured environment ofthe sealtoor. mulliresolution pyramidal blending of images to form a romposite senmles mosaic without blurring or loss of detail new image borders. io thw pnper we highlight the mathematicnl formulation behind ench uf these algorithms using consistent noration and n unified framework. We depict some of our M A T L A B toolbox results nith an nssorlment of underwater imagery. Index Tu"-Image processing, mosnicking. MATLAA. 1. I N I K 0 I ) U C I ' I O N NE aspect of our research here at the Deep Submergence 0 Laboratury of the Wood, Hulr Oceanographic Institution IS optical image processing and mosaicking. Pruductiun uf the high resolution, large area imagery of the seafloor required by our end user coniniunity. (geologists. biologists. archaeoiugi,ts.. . ) uould be impossible uere it not for the technique of pliu.tomosaicking whereby a large are3 image is assimilated from tens to hundmds of smaller-foutprint. overlapping images. The propagation of visible light underuater sufkrs from rapid attenuation and scattering. which in combinatiiin uith a limited camera-to-light separation akaiiable on most vehicle platforms. places strung constraints on the abilit) to iniage large a r e a uf the seaRwr optically. Special consideration must he giwn tu the unique cunstraints ssuciated with the underwater environment.Our research in underuater image processing and photumoraicking deals explicitly with these constraints. In this paper we cover Lome successtul techniques in the literature for image processing and mosaicking uhich have k e n extended to deal with unique peculiarities of the underuater enkirunment. The algorithms and order of sections presented in this paper arc:contrast limited adaptive histogram specification (CLAHS) to deal with the inherent nonuniform lighting in underwater imagely.Fourier based methods for scale, rotation, and translation recovery which provide robusmess against dissimilar image regions.. local normalized correlation for image registration to handle the low-feature, unstructured environment of the seafloor.. multiresolution pyramidal blending of images to form a composite seamless mosaic without blurring or loss of detail near image borders. As MATLAB has ...

Large-area visually augmented navigation for autonomous underwater vehicles

“…This technique was originally developed for multi-sensor fusion of optical and infrared imagery and has been successfully applied underwater in at least one known structure-from-motion (SFM) application [97]. Meanwhile, in [149] a technique for underwater imagery is developed by first preprocessing using adaptive histogram specification [37,182] (Fig. 1-7) followed by gray-level thresholding to detect and discount shadow regions in an attempt to account for lighting variations before attempting image registration.…”

Hacia la Navegación Visual de un Vehículo Autónomo Submarino en Áreas con Posidonia Oceanica

“…La figura 1 muestra una imagen extraída de una de las secuencias grabadas por el SPARUS II en el norte de Mallorca, con presencia de Posidonia y la misma imagen después de haber aplicado los 4 filtros. Adicionalmente, se ha aplicado Contrast Limited Adaptive Histogram Specification (CLAHS) (Eustice et al, 2002) en alguna secuencia de vídeo, solo para pares de imágenes que potencialmente puede cerrar un bucle. Este método de mejora de contraste está incluido en la libreria UWIT (Underwater Image Toolbox for Optical Image Processing and Mosaicking in MATLAB) (Eustice et al, 2002), y está diseñado para tratar la falta de uniformidad en la iluminación, efecto inherente a la toma de imágenes en entornos marinos.…”

“…Adicionalmente, se ha aplicado Contrast Limited Adaptive Histogram Specification (CLAHS) (Eustice et al, 2002) en alguna secuencia de vídeo, solo para pares de imágenes que potencialmente puede cerrar un bucle. Este método de mejora de contraste está incluido en la libreria UWIT (Underwater Image Toolbox for Optical Image Processing and Mosaicking in MATLAB) (Eustice et al, 2002), y está diseñado para tratar la falta de uniformidad en la iluminación, efecto inherente a la toma de imágenes en entornos marinos. La librería UWIT está especialmente enfocada a mejorar diversos aspectos en la construcción de foto-mosaicos, siendo el uso de CLAHS en las imágenes implicadas uno de los puntos claves en el registro de imágenes que cierran bucles.…”

LAPMv2: An improved tool for underwater large-area photo-mosaicking

“… Multi-frequency-bands blending: also known as multiresolution blending [3], [6], [21], this method consists of decomposing the images into a set of frequency bands and to blend each band over a different spatial range. The high frequencies are blended over a small range, while the low frequencies are blended over a large range [21]- [23].…”