The Colour Correction Of Underwater Target Optical Images Based On Beer’s Law

For marine sciences sometimes there is a need to perform underwaterphotography. Optical properties of light cause severe quality problems for underwaterphotography. Light of different energies is absorbed at highly different rates underwater causing significant bluishness of the images. If the colour dependent attenuationunder water can be properly estimated it should be possible to use computerizedimage processing to colour correct digital images using Beer’s Law.The first step is image preprocessing. All standard digital cameras have built inwhite balancing and colour enhancement functions designed to make the images asaesthetically pleasing as possible, and impact our methods. These functions can inmost cameras not be switched off and the algorithms used are proprietary andundocumented. However, these enhancement functions can be estimated. Applyingtheir reverse creates unenhanced images and eliminating the effects of enhancementfunctions for color restoration algorithm.In this thesis studiesthe colour correction methods based on Beer’s Law, thebasic process is to calculate the diffuse attenuation coefficient which put it into Beer’sLaw, derived the colour restoration algorithm. This factor can be obtained frommultispectral or hyper-spectral data. The main work is get better results improving thealgorithms, so that it is more accurate and economical.First, through the construction of the bottom compensation function, it can bebetter used at bottom reflection is too large or too small the colour correctionalgorithm based on Beer’s Law.Second, the improved methods for color correction of underwater picture ignorethat a will be scattering the light, and they only care the absorption of light, thus willcome out imprecise result. If we want to get a more accurate attenuation coefficient,we must compensate the loss of light caused by scattering. In order to improve the accuracy of the attenuation coefficient, we use Monte Carlo methods to simulate thetransfer process of multiple scattering of light in turbid water. And using Monte Carlomethod to calculate the irradiance at a distance z, and using this irradiance tocompensate the loss of light caused by scattering, we can get a more accurateattenuation coefficient. Then we can get a higher quality image after compensating theloss of light caused by scattering. Using Monte Carlo calculate the diffuse attenuationcoefficient have a low speed of convergence, but this method can be used in variouswater areas which have high concentration of solutes.Finally, the above improvements of methodbase on the RGB color imagesdirectly recover method. In this way less the number of spectral channels are neededto restore the color of images, so the effect is not ideal. To take advantage of morenumber of spectral channels, and to improve the recovery effect of the algorithm, heregives another method whichuse spectrometer to discrete spectral data into dispersedata, and produce a pseudo-hyperspectral image from the RGB image which capturedby the camera.Finally, the images can be weighted together in the proportions neededto create new correct RGB images. This method is somewhat computationallydemanding but gives very encouraging results. This method requires a great deal ofcomputing, but it can give us a very good recovery results.The algorithms and applications presented in this thesis show that automaticcolour correction of underwater images can increase the credibility of data takenunderwater for marine scientific purposes.And our algorithms and applications cancontribute to the cause of marine research.