MULTIMODAL APPROACH FOR DIGITAL IMAGE RESTORATION Seryazi Andrew Reg

MULTIMODAL APPROACH FOR DIGITAL IMAGE RESTORATION

Seryazi Andrew
Reg. No: 2017/HD05/2145U
Department of Computer Science
School of Computing and Information Sciences, Makerere University
E-mail: [email protected], Tel: +256703736225

A Concept Paper Submitted to School of Graduate Studies for the Study Leading to the
Research Proposal in Partial Fulfilment of the Requirements for the Award of Master of Science in Computer Science of Makerere University.
OPTION: Computer Vision and image processing

August 2018

INTRODUCTION

Digital images are referred to as electronic snapshots of a scene, the image is composed of matrix elements in a grid formation known as pixels. Each pixel holds a value which is quantized to represent the tone at a particular point. Digital images are captured in various fields which include photography, medical imaging, remote sensing, microscopy, astronomy and many others.
The image restoration process uses a priori knowledge of the degradation phenomenon 18. Restoration formulates and accesses the objective criteria of goodness. The imperfection in image quality can be modelled as noise or blur or a degradation function. Most images are more or less blurry. This is because there is too much interference in the camera as well as in the environment. Blurring of an image is caused by several factors which include movement during the capture process, network interference, wide angle lens, long exposure times and many others.
Digital image restoration may be visualized as a process in which we try to obtain an approximation to f(x, y) 6.
Images are affected by noise which is referred to as unwanted variation in the image 7. It normally leads to drastic change in visibility of an image. Digital images related to digital signals are usually corrupted by many types of noise like Inverse f Noise, Uniform noise, Salt and pepper noise (Impulse noise), Brownian Noise, Gaussian noise.
The field of digital image processing handles a wide range of activities which include extraction of features, assessment of images and restoration of images. Other related activities include the process of enhancement and filtering. Image restoration is one of the basic steps of processing that deals with making certain improvements in a digital image based on some predefined criteria. The main objective for restoration is to reconstruct an image that has been degraded based on some prior knowledge of the degradation phenomena of the image.
The process involved in restoration is objective in nature hence; it focuses at a specific goal like removal of blur by means of a deblurring function .The modalities that are used in the restoration of images are formulated in the spatial or frequency domain. Image restoration is based on probabilistic models of image degradation. Thus image restoration tends to make the images look better in appearance 91011.
The priorities of image restoration are the removal or reduction of degradations which are included during the acquisition of images for example Noise, pixel value errors, out of focus blurring or camera motion blurring using prior knowledge of the degradation phenomenon.
This means it deals with the modelling of the degradation and applying the process (inverse) to reconstruct the image. The image restoration has got a wide scope of usage.

RELATED WORK

Digital image denoising is a tremendous step in image processing, a new 13 image restoration strategy centred on Curvelet coefficients is applied. The noise in the image is processed through Curvelet thresholding, well as, the index set is preserved by the curvelet coefficients whose capacity level is higher than the thresholding level. Another similar image is obtained by applying the index set to the image difference between the noisy image and the reconstructed image through thresholding.
Image degradation is associated with many factors 5. Firstly they supply a short overview for the optical thought of defocused image, then discuss items of defocusing and presenting an effective approach to formulate the point spread perform (PSF) of defocus. Using the Gaussian model and degradation of defocus in parameter estimation, a new method is recommended to reconstruct defocused image, founded on Lucy Richardson Algorithm and blended with Wiener Adaptive filtering disposing of the noise. The results demonstrate that the new system can obtain fair recovery output.
Here 3 a regularized anisotropic diffusion filter was used to revive. The filtering modality displayed good posedness and great maintenance of edges. To evaluate its effectiveness in analysing the Rician noise, the PSNR and MSSIM metrics are adopted for the initial time. The output from the unreal and real knowledge demonstrated easier performance of the offered filters.
The research 42 presents a modified homomorphic deconvolution that is utilized to strengthen the quality of clinical ultrasound images. The deconvolution performs the homomorphic filtering centered on the estimation of the factor-unfold perform (PSF). The application of a non-regional way (NL-method) algorithm makes PSF estimation extra distinct for rejecting the White-Gaussian noise (WGN). They validate the system for exceptional radiofrequency (RF) images with resolution growth.
This research 12 reveals a new Bayesian approach for the restoration of blurred and noisy images. The Bayesian process depends on snapshot priors which encapsulate prior image abilities and avert the in poor health posedness of image restoration problems. They utilize a spatially various snapshot prior using a gamma-normal hyper prior distribution on the regional precision parameters. The proposed restoration manner is when compared with other photo restoration tactics, demonstrating its expanded performance.
.

PROBLEM STATEMENT
Digital image restoration is a new technology that has been embraced to recover images from the original image that has been degraded with prior knowledge of the degradation function.
There are still so many gaps and imperfections in the quality of images recovered depending on the approach or technique used.
There is still alot of work to be done in the field of digital image restoration to address these problems involved in the restoration process.
For the implementation of image restoration algorithms, the most important hindrance is to get well the degraded snapshot to a larger extent. It’s indispensable that resultant picture acquired after applying a restoration algorithm must be close to the customary snapshot. My proposed research and study would be on a multi-modal restoration approach or technique to address the digital image restoration imperfections.
Some research has been done on single techniques, however no research has been done on using a multi-modal approach.

Main Objective
The main objective of the study is to come up with a multi modal approach for digital image restoration.

Specific Objectives
Specific objectives of the study include the following:
1. Reviewing literature on digital image restoration.
2. Evaluating the performance of the different digital image restoration techniques.
3. To develop a multi-modal algorithm for better performance and output.

The anticipated outcomes resulting from this research are:
This research and study will provide a scientific contribution in the field of digital image processing but in particular the digital image restoration field by providing a better solution on how to restore degraded images with less imperfections.
The research provides a new approach in restoring digital images that have been degraded due to environmental and technical conditions.
This study will provide a base line for further investigation into digital image restoration for different applications.

Methodology
I propose to use a multimodal approach for digital image restoration. The first strategy is to scientifically evaluate and compare the results of the various filters to the original degraded snapshot. The second strategy is to fuse the three best evaluated filters and come up with a new algorithm with greater performance and superior output.
The experimental results will scientifically be analysed and studied to ensure that they achieve the goals and objectives of the research.
Most of the experiments and testing will be performed using digital image processing software like Mat Lab and Jupyter Notebook.

References

1 Dong-Dong Cao, Ping Guo, “Blind image restoration based on wavelet analysis,” IEEE, Machine Learning and Cybernetics, pp. 4977 – 4982, 2005.
2 Tzikas, D.G., Likas, A.C., Galatsanos, N.P, “Variational Bayesian Blind Image Deconvolution with Student Priors,” IEEE, Image Processing, pp.109- 112, 2007.
3 Zhang X. F, Ye H, Tian W.F, Chen W.F, “Denoising DWI Based on Regularized Filter,” IEEE, pp. 120-121, 2007.
4 Oleg Michailovich, Allen Tannenbaum, “Blind Deconvolution of Medical Ultrasound Images: A Parametric Inverse Filtering Approach,” IEEE, Image processing, pp. 3005-3019,2007.
5 Wei-Wen Wu, Jin-HuiZhong, Zhi-Yan Wang, “A new method for restoration of defocused image ,”IEEE,Machine Learning and Cybernetics (ICMLC),pp.2402 – 2405,2010.
6 Er.Jyoti Rani, Er.Sarabjeet Kaur. Volume 4, Issue 1, January 2014.International Journal of Advanced Research in Computer Science and Software Engineering,.
7 Er. Priya Tiwari, Dr. Naveen Dhillon and Er. Kuldeep Sharma.Vol.3, no. 4, 1.July 2013.Analysis of Image Restoration Techniques for Developing Better Restoration Method. 8 Birwa Raja Mohapatra, Ansuman Mishra, Sarah KumarRout.Vol.2,No.3, March 2014.A Comprehensive Review on Image Restoration Techniques.
9 Mathri thakur, Shilpa Chadury.Volume 14 Number 4 – Aug 2014.Image Restoration Based on Deconvolution by Richardson Lucy Algorithm. International Journal of Engineering Trends and Technology (IJETT).
10 R.Pushpavalli1 and G.Sivarajde2.Vol.4, No.2, April 2013.A hybrid filtering technique for eliminating uniform noise and impulse noise in digital image. Signal ; Image Processing. An International Journal (SIPIJ)
11 Giacomo Boracchi and Alessandro Foi. Vol. 21, no. 8, August 2012. Modeling the Performance of Image Restoration from Motion Blur. IEEE Transaction on image processing.
12 Rafael C.Gonzalez, Richard E. Woods, Steven L. Eddins, “Digital Image Processing Using MATLAB”, pp.499, 2009. Mateos, J., Bishop, T.E., Molina, R., Katsaggelos, A.K., “Local Bayesian image restoration using variational methods and Gamma-Normal distributions,” IEEE, Image Processing (ICIP), pp.129 – 132, 2009.
13 Qianzong Bao, Qingchun Li, “Image restoration with significant Curvelet coefficients index setconstrains”,IEEE,Information Theory