Description:HYBRID WCID APPROACH FOR ENHANCEMENT OF UNDERWATER IMAGES FOR CLASSIFICATION IMAGES
Field of Invention
It is required to reduce the volume of the image without effecting the quality and information contained in several communication systems. Hence, image compression without loss in the information which is termed as lossless image compression is considered to be a challenging task. Along with the high standard compression technique, the information contained image also needs to be protected with significant encoding and decoding techniques. In medical field, the data related to a particular disease is growing day by day. In the process of automation of diagnosis, it is required to sort out and arrange the images of several subjects. Hence, classification and clustering of images is important for proper analysis in short span of time. At times it is necessary to transfer the images in high density medium like under-water. This is more likely in the case of underwater monitoring through capturing images. Image enhancement techniques are often took failure path while applied to such applications.
Background of the Invention
There are a plethora of spatial domain techniques that can remove the aforementioned degradations in image quality. Iqbal et al. presented a colour equalisation approach to improve the quality of underwater photographs by correcting for the colour cast problem. The method entails determining the dominant colour plane in the RGB colour space and then using that plane's scaling factor to normalise the remaining colours. A major drawback of blind coloure qualisation is the degradation it causes to the image's colour quality. Beer's law, which is used to correct the pixel intensity by a calculation of the quantity of light absorption in water, is another technique for reducing colour cast. Assuming that everything in the scene is at the same depth, this approach determines the wavelengths that are missing. While this method does enhance image quality, it relies on the assumptions of a uniform medium and the same depth for all objects in the scene, which may not always hold true. Furthermore, the colour components depend on scene depth, water molecule size, water medium density, and local temperature, making accurate calibration of the enhancement settings essential. As was said before, water's ability to absorb light is a primary contributor to blurriness. Underwater vehicles use artificial lighting to illuminate the environment for the purpose of capturing images, despite the fact that this creates shadows due to the lighting's lack of uniformity. Methods are established based on the physics of image creation to combat the negative impacts of absorption and backscattering. To counteract the physical consequences of image quality loss, Iqbal et al. suggested an algorithm based on a set of photos taken through a polarizer at various angles.
An alternate strategy in these techniques involves enhancing lighting and sensing devices to illuminate an object before snapping a photograph. Using a lens filter is another common approach to this problem; by reducing the absorption and backscattering effects, the image quality is improved. Since the lens filter blocks some light, this method is more suited to improving terrestrial images. By comparing the historical colour data with the current one, Markov random fields can also be used for colour correction. The premise on which these approaches rest is that nearby images contain scenes that are very similar to the current one.

It is necessary to lower the volume of the image while maintaining the quality and information included in various communication systems. As a result, image compression without information loss, also known as lossless image compression, is regarded as a difficult undertaking. Along with a high quality compression approach, the image's information must be preserved by considerable encoding and decoding procedures. The aforementioned degradations in image quality can be removed using any number of spatial domain techniques (US10552663B2). Iqbal et al. presented a colour equalisation approach to improve the quality of underwater photographs by correcting for the colour cast problem. This technique uses the dominant colour plane's scale factor in the RGB colour space to normalise the other colours. Blind colour equalisation, on the other hand, has been shown to significantly diminish image quality. Beer's law, which is used to correct the pixel intensity by calculating the quantity of light absorption in water, is another technique for reducing colour cast. This technique for calculating missing wavelengths assumes that everything in the scene is at the same depth. The image quality is improved by this method, but it may be unrealistic to assume that all objects in a scene have the same depth. Accurate calibration of the enhancement parameters is also required because the colour components are affected by variables such as the scene's depth, the size of water molecules, the density of the water medium, and the local temperature. In practise, it is crucial for extended mission endurance of AUVs to make efficient use of on-board electrical energy; as a result, on-board lights are often flashed just at the moment of the image acquisition, resulting to insufficient information about the neighbouring scene. In a heterogeneous undersea environment, the previous data may not be equivalent to the current one, making statistical methods of improvement impractical (US10223610B1).
Summary of the Invention
The processing techniques for images under water are entirely different from that of the free space. In order to accomplish the task of efficient recognition of the image under water, two different techniques are studied. The first refers to the WCID and the latter is the stationary wavelet transform. Underwater and foggy photographs benefit greatly from the suggested hybrid WCID method's ability to efficiently restore colour and remove hazes. Aqua formation circumstances change throughout time in the actual world as a result of things like weather. Because of these deviations, estimating the rate at which energy is lost is imprecise. Aside from this, the suggested hybrid WCID yields satisfactory improved images under any other stipulations. Therefore, this method is particularly useful for analysing and monitoring the health and growth rate of prawns and fish in aquaculture. The enhancement of underwater image is done using Stationary wavelet transform (SWT). This technique restored the degraded underwater image very effectively. The tint in the degraded image is reduced by white balancing which uses mean filtering. The contrast of image is increased by color correction using histogram equalization.
Brief Description of Drawings
Figure 1: Underwater Image capturing model
Figure 2: Architecture of the proposed method

Detailed Description of the Invention
Dehazing is one of the best techniques for improving blurry photos of low quality. While this approach does improve image contrast, the colour cast issue caused by the wavelength mismatch is not fixed. However, while histogram equalisation also improves image contrast, it does nothing to fix issues with colour cast or dispersion. In order to improve image contrast and eliminate colour scatter and cast, this work introduces a novel hybrid WCID method. According to the image capture model depicted in Figure 1, where R represents the camera's capturing range, D represents the depth to which the camera's capturing region extends below the surface. The image is formed as the reflected light from the item x travels a distance d(x). Light absorption during photography causes a scattering of colours in the final image. Accurate depth calculation from each pixel in the image is required to correct for colour scatter and cast. Underwater environments often require an additional light source (L) to make up for natural light's limitations. In WCID, restoring the cast is the primary method for performing inverse compensation after scatter has been removed. WCID uses two criteria to determine pay: Both R and L are measures of energy loss at a distance. Image enhancement is a sort of pre-processing that is performed before an image is used. Colour, grayscale, underwater, satellite, etc. images can all benefit from this technique.
In these strategies, an alternative strategy involves boosting lighting and sensor devices to highlight an item before taking an image. Another frequent solution to this problem is to use a lens filter; by lowering absorption and backscattering effects, image quality is improved. Because the lens filter reduces light, this method is more suited to improve terrestrial photos. Markov random fields can also be utilized for color correction by comparing historical and present color data. These approaches are based on the notion that nearby photos contain scenarios that are quite similar to the current one.
A colour equalisation strategy to improve the quality of underwater images by adjusting for the colour cast problem. This technique normalizes the remaining colors by using the scale factor of the dominating color plane in the RGB color space. Blind color equalisation, on the other hand, has been proven to reduce image quality dramatically. Another strategy for eliminating color cast is Beer's law, which is used to correct pixel intensity by calculating the amount of light absorption in water. This method for determining missing wavelengths is based on the assumption that everything in the scene is at the same depth. This method improves image quality, but it may be unreasonable to presume that all objects in a scene have the same depth. Calibration of the boost is precise.
This method calculates the wavelengths that are absent by assuming that everything in the scene is at the same depth. While this method improves image quality, it is based on the assumptions of a uniform medium and uniform depth for all objects in the scene, which may not always be true. Furthermore, color components are affected by scene depth, water molecule size, water medium density, and local temperature, necessitating precise calibration of the enhancement parameters. As previously stated, the propensity of water to absorb light is a major contribution to blurriness. Underwater vehicles employ artificial illumination to illuminate the area in order to capture photographs, despite the fact that the lighting is not consistent, resulting in shadows.
Histogram Equalisation, the Dark-channel previous method, Wavelength Compensation, and Image Dehazing (WCID) are all methods that can be used to enhance underwater photographs. Image contrast can be improved via histogram equalisation, however it cannot make up for the issue of light scattering. The deteriorated image's light scattering issue is fixed using the dark channel previous approach. This issue occurs when the object's scene has lighting conditions that are too close to those of the background. So this method is not preferable for enhancing the underwater image. Wavelength Compensation and Image Dehazing (WCID) technique, dehazing algorithm is combined with the wavelength compensation. The dark prior model is used to estimate the distance between camera and the target (object). Dehazing algorithm is used to remove the color scattering. Next photography scene depth is estimated from the energy of each wavelength in the background of an image. Restoring the distortion from color cast is performed by reverse compensation. Wavelength compensation and image dehazing algorithm (WCID) overcomes all the problems related to color change effect and light scattering. Unsupervised Color Correction Method (UCM) is based on color balancing and contrast correction of HIS (Hue Saturation and Intensity) color model and RGB color model. This model provides color balance in an image. It improves the illumination with an increase in actual color by removing the color cast.
Typically, d(x) between two images requires a distance of two images. However, under the sea, the hazes rise proportionally with distance travelled. Therefore, a single image is adequate in this WCID for distance estimation. The details of submerged objects are retrieved using the stationary wavelet transform. There are three crucial steps in this process. Figure 2 depicts the procedure to follow.
First step is White balancing is the process of modifying the colour components to remove the bluish and reddish tint and the unrealistic colour casts so that whites seem white in the image. When adjusting the white balance of a photo, it's important to take the image's colour temperature into account. White light can be either cold or warm, depending on its colour temperature. Mean filtering is used to achieve proper white balance. It's one of the best ways to capture sea life in pictures. Second one is Color correction Generally, images are degraded due to poor contrast and brightness. Color correction is done by adjusting the contrast adjustment with the help of histogram equalization. Third one is Image Fusion using Stationary Wavelet Transform: In machine vision, image does not contain all the information. So it is preferable to fuse to images.
Image fusion is a process of combining two or more images into a single image for extracting the important features of each single image. Image fusion is done using spatial fusion. Types of spatial fusion are Simple Averaging is performed as the pixel values of two images is taken and added and then Average is obtained by dividing this sum by 2. After that Select Maximum is achieved as the focus of the image depends on the pixel values. If the pixel values are of higher value then, the focus is more. In this algorithm, the in-focus regions are choosen using the higher pixel values which results in highly focused image. The greatest pixel value is assigned to the corresponding pixel
Image fusion is done using many transforms like discrete wavelet transform and Pyramid Transform. Multi-scale transforms use discrete wavelet transform or pyramid transform for representing the source image at multi-scale. The disadvantage of pyramid transform includes lack of flexibility and blocking effects. To overcome these problems, DWT approach is considered. But, discrete wavelet transform (DWT) suffers from poor directionality and lack of shift invariance. To avoid these disadvantages a Dual Tree Complex Wavelet Transform (DTCWT) is used.
4 Claims &2 Figures , Claims:The scope of the invention is defined by the following claims:

Claim:
The Design of a Hybrid WCID approach for enhancement of underwater images for Classification images comprising the steps of:
a) Designed a technique that has to analyse the images in terms of colour and contrast.
b) Adopted a method for enhancing the quality of the underwater image
c) Design architecture for describing the classification of images step by step.
2. The Design of a Hybrid WCID approach for enhancement of underwater images for Classification images as claimed in claim1, the approaches Dehazing and Histogram equalization are used for enhancing the blurred image and colour contrast.
3. The Design of a Hybrid WCID approach for enhancement of underwater images for Classification images as claimed in claim1, a technique WCID is developed for colour scatter and cast problems and to enhance under water images.
4. The Design of a Hybrid WCID approach for enhancement of underwater images for Classification images as claimed in claim1, Adopted a method of Dual Tree Complex Wavelet Transform (DTCWT) s designed.