A Preprocessing Framework for Underwater Image Denoising Essay
Sorry, but copying text is forbidden on this website!
A major obstacle to underwater operations using cameras comes from the light absorption and scattering by the marine environment, which limits the visibility distance up to a few meters in coastal waters. The preprocessing methods concentrate on contrast equalization to deal with nonuniform lighting caused by the back scattering. Some adaptive smoothing methods like anisotropic filtering as a lengthy computation time and the fact that diffusion constants must be manually tuned, wavelet filtering is faster and automatic. An adaptive smoothing method helps to address the remaining sources of noise and can significantly improve edge detection.
In the proposed approach, wavelet filtering method is used in which the diffusion constant is tuned automatically. Keywords: underwater image, preprocessing, edge detection, wavelet filtering, denoising.
The underwater images usually suffers from non-uniform lighting, low contrast, blur and diminished colors. A few problems pertaining to underwater images are light absorption and the inherent structure of the sea, and also the effects of colour in underwater images. Reflection of the light varies greatly depending on the structure of the sea. Another main concern is related to the water that bends the light either to make crinkle patterns or to diffuse it. Most importantly, the quality of the water controls and influences the filtering properties of the water such as sprinkle of the dust in water. The reflected amount of light is partly polarised horizontally and partly enters the water vertically. Light attenuation limits the visibility distance at about twenty meters in clear water and five meters or less in turbid water. Forward scattering generally leads to blur of the image features, backscattering generally limits the contrast of the images. The amount of light is reduced when we go deeper, colors drop off depending on their wavelengths. The blue color travels across the longest in the water due to its shortest wavelength. Current preprocessing methods typically only concentrate on local contrast equalization in order to deal with the nonuniform lighting caused by the back scattering.
II. UNDERWATER DEGRADATION
A major difficulty to process underwater images comes from light attenuation. Light attenuation limits the visibility distance, at about twenty meters in clear water and five meters or less in turbid water. The light attenuation process is caused by the absorption (which removes light energy) and scattering (which changes the direction of light path). Absorption and scattering effects are due to the water itself and to other components such as dissolved organic matter or small observable floating particles. Dealing with this difficulty, underwater imaging faces to many problems: first the rapid attenuation of light requires attaching a light source to the vehicle providing the necessary lighting.
Unfortunately, artificial lights tend to illuminate the scene in a non uniform fashion producing a bright spot in the center of the image and poorly illuminated area surrounding. Then the distance between the camera and the scene usually induced prominent blue or green color (the wavelength corresponding to the red color disappears in only few meters). Then, the floating particles highly variable in kind and concentration, increase absorption and scattering effects: they blur image features (forward scattering), modify colors and produce bright artifacts known as “marine snow”. At last the non stability of the underwater vehicle affects once again image contrast.
To test the accuracy of the preprocessing algorithms, three steps are followed.
1) First an original image is converted into grayscale image. 2) Second salt and pepper noise added to the grayscale image. 3) Third wavelet filtering is applied to denoise the image. Grayscale images are distinct from one-bit bi-tonal black-and-white images, which in the context of computer imaging are images with only the two colors, black, and white. Grayscale images have many shades of gray in between. Grayscale images are also called monochromatic, denoting the presence of only one (mono) color (chrome). Grayscale images are often the result of measuring the intensity of light at each pixel in a single band of the electromagnetic spectrum and in such cases they are monochromatic proper when only a given frequency is captured. Salt and pepper noise is a form of noise typically seen on images. It represents itself as randomly occurring white and black pixels. An image containing salt-and-pepper noise will have dark pixels in bright regions and bright pixels in dark regions. This type of noise can be caused by analog-to-digital converter errors, bit errors in transmission. Wavelet filtering gives very good results compared to other denoising methods because, unlike other methods, it does not assume that the coefficients are independent.
III. A PREPROCESSING ALGORITHM
The algorithm proposed corrects each underwater perturbations sequentially. addressed in the algorithm. However, contrast equalization also corrects the effect of the exponential light attenuation with distance.
B. Bilateral Filtering
Bilateral filtering smooth the images while preserving edges by means of a nonlinear combination of nearby image values. The idea underlying bilateral filtering is to do in the range of an image what traditional filters do in its domain. Two pixels can close to one another, occupy nearby spatial location (i.e) have nearby values. Closeness refers to vicinity in the domain, similarity to vicinity in the range. Traditional filtering is a domain filtering, and enforces closeness by weighing pixel values with coefficients that fall off with distance. The range filtering, this averages image values with weights that decay with dissimilarity. Range filters are nonlinear because their weights depend on image intensity or color. Computationally, they are no more complex than standard nonseparable filters. So the combination of both domain and range filtering is known as bilateral filtering.
A. Contrast equalization
Contrast stretching often called normalization is a simple image enhancement technique that attempts to improve the contrast in an image by ‘stretching’ the range of intensity values. Many well-known techniques are known to help correcting the lighting disparities in underwater images. As the contrast is non uniform, a global color histogram equalization of the image will not suffice and local methods must be considered. Among all the methods they reviewed, Garcia, Nicosevici and Cufi  constated the empirical best results of the illuminationreflectance model on underwater images. The low-pass version of the image is typically computed with a Gaussian filter having a large standard deviation. This method is theoretically relevant backscattering, which is responsible for most of the contrast disparities, is indeed a slowly varying spatial function. Backscattering is the predominant noise, hence it is sensible for it to be the first noise
Anisotropic filter is used to smoothing the image. Anisotropic filtering allows us to simplify image features to improve image segmentation. This filter smooths the image in homogeneous area but preserves edges and enhance them. It is used to smooth textures and reduce artifacts by deleting small edges amplified by homomorphic filtering. This filter removes or attenuates unwanted artifacts and remaining noise. The anisotropic diffusion algorithm is used to reduce noise and prepare the segmentation step. It allows to smooth image in homogeneous areas but it preserves and even enhances the edges in the image.
Here the algorithm follow which is proposed by Perona and Malik . This algorithm is automatic so it uses constant parameters selected manually. The previous step of wavelet filtering is very important to obtain good results with anisotropic filtering. It is the association of wavelet filtering and anisotropic filtering which gives such results. Anisotropic algorithm is usually used as long as result is not satisfactory. In our case few times only loop set to constant value, to preserve a short computation time.
For this denoising filter choose a nearly symmetric orthogonal wavelet bases with a bivariate shrinkage exploiting interscale dependency. Wavelet filtering gives very good results compared to other denoising methods because, unlike other methods, it does not assume that the coefficients are independent. Indeed wavelet coefficients in natural image have significant dependencies. Moreover the computation time is very short.
IV. EXPERIMENTAL SETUP AND EVALUATION
To estimate the quality of reconstructed image, Mean Squared Error and Peak Signal to Noise Ratio are calculated for the original and the reconstructed images. Performance of different filters are tested by calculating the PSNR and MSE values. The size of the images taken is 256×256 pixels. The Mean Square Error (MSE) and the Peak Signal to Noise Ratio (PSNR) are the two error metrics used to compare image compression quality. The MSE represents the cumulative squared error between the compressed and the original image, whereas PSNR represents a measure of the peak error. The lower the value of MSE, the lower the error. In Table 1, the original and reconstructed images are shown. In table 2, PSNR and MSE values are calculated for all underwater images. PSNR value obtained for denoised images is higher, when compare with salt and pepper noise added images. MSE value obtained for the denoised images has lower the error when compared with salt and pepper noise added images. e
D. Wavelet filtering
Thresholding is a simple non-linear technique, which operates on one wavelet coefficient at a time. In its most basic form, each coefficient is thresholded by comparing against threshold, if the coefficient is smaller than threshold, set to zero; otherwise it is kept or modified. Replacing the small noisy coefficients by zero and inverse wavelet transform on the result may lead to reconstruction with the essential signal characteristics and with the less noise. A simple denoising algorithm that uses the wavelet transform consist of the following three steps, (1) calculate the wavelet transform of the noisy image (2) Modify the noisy detail wavelet coefficients according to some rule (3) compute the inverse transform using the modified coefficients. Multiresolution decompositions have shown significant advantages in image denoising.
best denoised image. In clearly, the comparisons of PSNR and MSE values are shown in Fig -1a and Fig -1b.
In this paper a novel underwater preprocessing algorithm is present. This algorithm is automatic, requires no parameter adjustment and no a priori knowledge of the acquisition conditions. This is because functions evaluate their parameters or use pre-adjusted defaults values. This algorithm is fast. Many adjustments can still be done to improve the whole pre-processing algorithms. Inverse filtering gives good results but generally requires a priori knowledge on the environment. Filtering used in this paper needs no parameters adjustment so it can be used systematically on underwater images before every pre-processing algorithms.
 Arnold-Bos, J. P. Malkasse and Gilles Kervern,(2005) “Towards a model-free denoising of underwater optical image,” IEEE OCEANS 05 EUROPE,Vol.1, pp.234256.  Caefer, Charlene E.; Silverman, Jerry. &Mooney,JonathanM,(2000) “Optimisation of point target tracking filters”. IEEE Trans. Aerosp. Electron. Syst., pages 15-25.  R. Garcia, T. Nicosevici, and X. Cufi. (2002) “On the way to solve lighting problems in underwater imaging”. In Proceedings of the IEEE Oceans 2002, pages 1018–1024.  James C. Church, Yixin Chen, and Stephen V., (2008) “A Spatial Median Filter for Noise Removal in Digital Images”, page(s):618 – 623. [45 Jenny Rajan and M.R Kaimal., (2006) “Image Denoising Using Wavelet Embedded anisotropic Diffusion”, Appeared in the Proceedings of IEEE International
Conference on Visual Information Engineering, page(s): 589 – 593.  Z. Liu, Y. Yu, K. Zhang, and H. Huang.,(2001) “Underwater image transmission and blurred image restoration”. SPIE Journal of Optical Engineering, 40(6):1125–1131.  P. Perona and J.Malik, (1990) “Scale space and edge detection using anisotropic diffusion,” IEEE Trans on Pattern Analysis and Machine Intelligence, pp.629-639.  Schechner, Y and Karpel, N., (2004) “Clear Underwater Vision”. Proceedings of the IEEE CVPR, Vol. 1, pp. 536-543.  Stephane Bazeille, Isabelle, Luc jaulin and Jean-Phillipe Malkasse, (2006) “Automatic Underwater image PreProcessing”, cmm’06 – characterisation du milieu marine page(s): 16-19.  Yongjian Yu and Scott T. Acton, (2002) “Speckle Reducing Anisotropic Diffusion”, IEEE Transactions on Image Processing, page(s): 1260-1270, No. 11, Vol.11.