Restoration of Obscured Images

Abstract

For the aim of exploring the deep undersea world map, a picture of high quality without interfering objects is preferred. However, within the water, the image quality tends to be hampered by light scattering, water density, and light attenuation. effects. Besides, the dynamic interference may affect the important underwater map. during this paper, we proposed a multi-step and all-around underwater image processing system, especially for the underwater images taken in succession to enhance the image quality, remove the dynamic interference, and reconstruct the image. The first step involves utilizing the dark channel approach together with the improved gray world algorithm for brightness adaptation and color correction. Initially, it identifies and removes a dynamic interference regarding image enhancement. Secondly, we applied an upgraded total variation model to patch the blank at the value of resolution. Finally, the super-resolution of the small print is realized by applying an improved BP network. After simulation and experiments, our system proved to realize ideal results of image enhancement and reconstruction.

Introduction

I. INTRODUCTION

The oceans are home to a variety of enigmatic and unidentified species as well as a vast supply of energy-producing resources, which contribute significantly to the sustainability of life on Earth [1]. Since the turn of the 20th century, high-tech maritime exploration initiatives have been undertaken globally [2]. Because of its ability to hold large amounts of data, vision technology has received a lot of attention [3]. In several undersea applications, including robotics [4], rescue operations, man-made structural inspection, ecological monitoring, tracking marine life [5], and real-time navigation systems [6], [7], researchers and scholars seek to obtain high-resolution underwater photographs. Academics and researchers are looking for high-resolution underwater photos. However, the underwater environment severely degrades image quality, causing problems that are easier to fix in terrestrial imaging. Images taken underwater consistently exhibit color casts, such as green-bluish hues, which are brought on by varying degrees of red, blue, and green light attenuation. Particles in suspension also have Images with a lot of blur and haze are caused by the fact that submerged objects absorb most photon energy and shift the direction of light before it reaches the camera [8]. Artificial light sources are widely employed in underwater imaging to extend their effective range. However, scattering and absorption have an impact on artificial light [9]. Simultaneously, an uneven lighting pattern is applied, leading to luminous areas in the center of the underwater picture and inadequate lighting in the direction of the boards [10Shadowing is one of the other phenomena that degrades quality. For underwater photographs to yield meaningful details, trustworthy methods for color correction, sharpening, and background scattering removal are therefore required. These are especially difficult because of the complicated underwater environment, where turbidity, light absorption, and scattering—all of which can vary widely cause lower-quality photographs.

II. UNDERWATER IMAGE ENHANCEMENT METHODS

We approach the handling of specific problems in specific images using different algorithms. Many things are not perfect in the images taken underwater like the color attenuation, dark channels, intensity of the image, and adequate light to capture the image. We propose to use different algorithms for different problems to get the desired results. Underwater image enhancement is crucial to improve the visibility and quality of images in aquatic environments, where factors like attenuation, color distortion, and poor lighting significantly degrade image quality. Several methods and techniques are employed for underwater image enhancement

A. Contrast Limited Adaptive Histogram Equalization

Contrast Limited AHE (CLAHE) is a kind of adaptive histogram equalization designed to lessen the issue of noise amplification by limiting contrast amplification. In CLAHE, the modification of discrepancy in the vicinity of a pixel value is given by the pitch of the metamorphosis function. The value of the histograms at this pixel value, which is connected to the value of the pixel in this picture, equals the pitch of the original CDF. CLAHE restricts amplification by cutting off or “clipping” the histogram, setting its highest point before computing the CDA.

Additionally, it limits the CDF's slope, which suggests the transformation function's slope. Therefore, it relies on the neighborhood window's size and how the histogram is normalized, or more specifically, its clip limit. The ensuing amplification is limited to no more than three or four times by common values.

B. Rayleigh Stretching and Averaging

Underwater image generally consists of bright and dark areas. Overall image discrepancy can be increased by the global stretching of the image. The discrepancy of the image needs to be increased to increase the brilliance of darker areas still, global stretching also leads to an increase in the brilliance of the bright areas, leading to the over-improvement of the bright areas. Too bright, performing in loss of details. The same case occurs when global embroidering is applied to the darker areas because it produces under-meliorated areas that reduce image details. The proffered system, which uses two nonidentical image contrasts, addresses these effects. The eidolon of the proffered system is to produce two images with nonidentical contrasts one as an under-meliorated image and another as an over-enhanced image. The ensuing expressway curtly explains the system of producing two images from a single channel in the RGB color model. These expressways are applied to other channels.

The two produced images are piled together, and the average value between the two images is calculated. The image and the S and V factors of the HSV color model are stretched within the dynamic range of 1 percent from the minimum to the outside values. The final image is attained by converting. After applying these processes to the corresponding channels, all lesser-stretched histograms are formulated to produce an image, and all upper-stretched histograms are also formulated to produce another image. This process produces two nonidentical images                          with nonidentical contrasts. The lesser region of the histogram produces an under-meliorated image, whereas the upper region produces an over-meliorated image.

C. Relative Global Histogram Stretching    

RGB channels. Astride the fact that the characterized histogram is ignored over different channels and images. Upon application of fixed values say 1,244, the resultant may stretch over both extremes, i.e., overstretch or understretch for specific color channels and spoil the original image specifics. It is to be observed that the law of propagation of light beneath water dictates we apply the contrast correction method for the modification of deformed images. Also, the distribution rule of histograms over colored channels viz RGB, notes the following observations in images that are from shallow water: Majority of shallow-water images, the red light histogram lies in values [50, 150], whereas, the G channel and B channel are identically concentrated in the range [70, 210].

This is an indication of the fact that stretching of histograms should be sensitive to all channels. The Adaptive parameter obtained for a stretch range: From the distribution in the histogram for various images from shallow water, it can be observed that the distribution over the RGB channel which is similar to the change in Rayleigh distribution, is a continuous probability distribution over positive random variables. Also, a point of interest is that the channel distribution shows normalcy, and its mode and midpoint almost being same. Hence, we take the mode value as a divider for the individual decision of intensity level extremes, that is maximum and minimum, for the input image in stretching of histogram. The images from underwater are influenced by many factors, the need to decrease the impact of a few extreme pixels on the stretching of the relative global histogram, usually the stretching range lies between 0.1 percent and 99.9 percent of the histogram. But, in case the histogram is not normally distributed, this method which removes equal pixels is not feasible. Therefore we partition over upper and lower rates of the intensity values to enable the calculation of the I [min] and the stretching of the histogram.

D.  Underwater Light Attenuation Prior

Light attenuation refers to reduced intensity of light due to scattering and absorption of light by particles underwater when it travels from one medium to another. Restoring a hazy Underwater image is a tedious task when done using computer vision.

Humans can rapidly perceive the scene depth of the underwater image with no additional information. The furthest point within the depth map resembling the original underwater image is commonly considered because of the background light. With the weakening of sunlight underwater, depending on the wavelength, the energy of red light is absorbed more than that of green and blue lights, the highest intensity difference between Red light and Green-Blue light is employed to determine the background light within the underwater image. it’s often considered a background light candidate. After examining an outsized number of underwater images, we discover the Underwater Light Attenuation Prior (ULAP), which is the difference between the most value of G-B intensity and therefore the value of R intensity in one pixel of the underwater image is extremely strongly associated with the change of the scene depth

In underwater photos, a few pixels in at least one color channel have an intensity that is almost zero. The low intensity might be attributed to the shadows. Independent wavelength assumption does cause issues, even when the Dark Channel assumption is correct. In many real-world scenarios, the red channel is either completely black or almost completely dark. In this case, DCP corrupts the transmission estimation. The absorption of light by the medium causes the red channel to become nearly nil, even in shallow seas. The data gleaned from the red channel can't be trusted. UDCP, which takes into account the green and blue channels rather than the red channel, is utilized to get around the aforementioned problem.

F. Maximum Intensity Projection
R channels of subaquatic cinema, Carlevaris- Bianco, set up an expressway wherein an excellent difference between the immersion of RGBwavelengths was discovered. supported this former data, calculation of the depth of the scene is administered. The proffered algorithm is therefore called. The variation between two extreme valuations of R- R-channel vehemence and BG-channel emphasis is assumed for medium-transmission estimation. The trial substantiated that MIP can reckon rough distance maps of submarine cinema. The proffered algorithm is constantly utilized for atmospheric-light estimation further. Wen proffered the bettered interpretation of MIP( i.e. new optical model( NOM)) to cipher the a

References

[1] Lu, T. Uemura, D. Wang, J. Zhu, Z. Huang, and H. Kim, “DeepSea Organisms Tracking Using Dehazing and Deep Learning,” Mob.Netw. Appl., Oct. 2018. [2] X. Guo, Y. Li, and H. Ling, “LIME: Low-Light Image Enhancement via Illumination Map Estimation,” IEEE Trans. Image Process., vol. 26, no. 2, pp. 982–993, Feb. 2017. [3] Shallow-Water Image Enhancement Using Relative Global Histogram Stretching Based on Adaptive Parameter Acquisition Dongmei Huang, Yan Wang1, Wei Song, Jean Sequeira, and Sébastien Mavromatis. [4] Single Image Haze Removal Using Dark Channel Prior Kaiming He, Jian Sun, and Xiaoou Tang. [5] Initial Results in Underwater Single Image Dehazing Nicholas Carlevaris

Copyright

Copyright © 2023 Dr. R. Senkamalavalli, Pavan C, Sameer Kulkarni, Sudeep D M, Tahreem Pasha. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.