H. Fan, J. Li, Q. Yuan, X. Liu, M. Ng (2019) This paper [1] quantitatively evaluates results of the different Image Processing techniques in the simulated experiments. In this research it was concluded that Gaussian noise, stripes and deadlines can be removed from any image by using LRMR technique. This method has achieved a promising result by taking into consideration the relation between spectral and spatial dimensions of HSI.

Reginald L. Lagendijk and Jan Biemond (2019) This book describes in detail about the various processes of the image restoration. It also describes importance of restoration of images [2]. The main focus of this book is on the improvement of medical images of any human being.

R. Wang, Wei Li, Rui Li , Liang Zhang (2019) Using Support Vector Machines (SVM), this research suggests describing the blur type category of digital images [3]. Every image is subjectedto one of 3 types of blur: motion, defocus, and haze. 35 blur features are created using transform domains and image spatial information in this proposed methodology.

T.Ince (2019) Presented a mixing prior model for the spatial and spectral regions that takes advantage of the variation in HISs in spatial and spectral regions [4].

S.Bourennane,C.Fossati, Lin (2018) Demonstrated the PARAFAC model, which developed a model using a dynamic multilinear algebra model known parallel analysis for finding, an efficient de-noising method for the suppression of Gaussian noise from HSIs in this study (PARAFAC)[5].

L.Zhang, Q. Yuan, and H. Shen (2018) An HSI de-noising technique based on a model SSAHTV (spectral-spatial adaptive total variation)in this research [6.]SSAHTV method noise reduction method utilises spectral noise variance and spatial information discrepancies in account.

L Mredhula, M Dorairangaswamy (2017) This article describe how signal features can be used to remove noise. The main aim is to work on variable parameters of image and noises. Here medium filters were employed to remove noises and also describes importance of AI in image processing [7]

D. Sangeetha and P. Deepa (2016) Determines edge detection with the latest technique. In paper edge detection algorithm is used.This architecture decreases processing time by 6.8% and uses very few resources for edge detection process[8].

Wei He, Liangpei Zhang, Huanfeng Shen, Qiangqiang Yuan, Hongyan Zhang,(2014) According to the research, throughout the acquisition process, hyper spectral images are commonly damaged by mixture of the numerous type noises, including deadlines, stripes, impulse noise, and other phenomena. In this work, novel HSI restoration technique based on low-rank matrix recovery is presented that substantially suppresses Gaussian noise, impulse noise, delays, and colors [9]. The low-rank attribute of hyper spectral imagery is explored by lexicographically structuring a patch of th e HSI into a 2-D matrix, which means that clean HSI patch can viewed as low-rank matrix. LRMR framework is then expanded to accommodate the HSI restoration task. To overcome "Go Decomposition" approach is to solve LRMR issue without mixed noise. The performance of the suggested LRMR-based HSI restoration strategy has validated by experiments using simulated or real world data situations.

Yazeed A. Al Sbou (2012) Presented neural network as a noise reduction efficient and robust tool. In this research the BPNN is used as a learning algorithm. This approach includes using both mean and median stastical functions for calculating the output pixels of the NN. This uses a part of distorted image pixels to generate the system training pattern. The output of the proposed approach provided a good image de-noising performance which exhibits promising results of the degraded noisy image in terms of PSNR, MSE and visual test.

Low Rank Matrix Recovery(LRMR) Digital image methodology plays a very important role within the investigation and rationalization of remotely perceived knowledge. Image restorations techniques ease in enhancing the visibility of any zero.5 or feature of the image by dominant the knowledge in many parts or properties. Image restoration improves the clarity of objects within the scene by increasing the shine feature between objects and their backgrounds. Image restorations unit usually conducted as a distinction stretch followed by a tonal restoration. Neural Network (NN) A NN may be a machine learning approach galvanized by within which the brain performs a selected learning task. A neural network is associate assessing style that consists of massively parallel relation of adaptation 'neural' processors. As a result of its parallel nature, it'll conduct calculations at a high ranking as compared to the classical techniques. As a result of its variable nature, it wills befits changes among the information and learn the characteristics of sign. Outcome from one node is delivered to more one among the network and thus the top product depends on the advanced correlation of all nodes.

Support Vector Machine(SVM) It’s basically a classification wherein the optimization criteria are the width of the margin between the classes, i.e., the vacant region surrounding the decision border specified by the distance to the nearest training patterns. Support vectors are what they're called. Support vectors modify prototypes, which is the main difference between SVM and traditional template matching algorithms. A choice limit distinguishes the classes. The minimal distance function does not define this decision threshold. Vapnik introduced the Support Vector Machine (SVM) idea. Given a finite quantity of training data, any learning system's purpose is to attain high generalisation performance. With no prior understanding of the data, support vector machines have demonstrated high generalization performance. SVM's primary concept is to map the input data. On to 3-D feature space discover a separation hyper plane that is nonlinearly related to the input space in the feature space which has the max margin between the two classes. SVM is a maximal margin hyperplane in feature space based on kernel functions. As a result, the data space has a nonlinear boundary. Using kernel functions in the input space, the best separation hyper plane can be calculated without any calculations in the higher dimensional feature space. The following are some of the most often utilized kernels:

1. Liner Kernel- A Linear Kernel is employed when the data is Linearly separable, that is, it can be split using a single Line. It is one of the most often used kernels. It is commonly used when a data set has a significant number of characteristics.

2. Polynomial Kernel- Polynomial kernels are kernel functions commonly used with support vector machines and other kernelized models to express the similarity of vectors in a feature space over polynomials of the original variables, allowing non-linear models to be learned. GEOMETRICAL REPRESENTATION OF SVM MARGIN Figure. 

1. Geometrical representation of SVM margin SVM ALGORITHM

i. . Define a hyperplane that is optimal.

ii. Extend previous definition to include issues that are not linearly separable.

iii. Transform data into a high-dimensional space where linear decision surfaces may be used to categorize it more easily.