Regularizing of Inverse Problem

[Link]
(P03NK21S0404)
Under the Supervision of
Dr Harina P. Waghamore

DEPARTMENT OF MATHEMATICS
Bangalore University, Bangalore - 560056

[Link] Regularizing of Inverse Problem 0 / 22

Table of Content

Introduction
Definitions
Concepts
Applications
Conclusion
Bibliography

[Link] Regularizing of Inverse Problem 1 / 22

Introduction

Regularization in the context of inverse problems refers to a set of

techniques used to stabilize and find a stable solution for ill-posed
problems. Inverse problems involve determining the causes or inputs
from observed outcomes or outputs, often with noise or incomplete
information. If you have ever driven a car, watched the weather
channel, had a CAT scan or MRI then it is likely that your life has
benefited in some way from the solution of inverse problems. Solving
an inverse problem is the process of recovering some hidden
information such as a set of parameters from indirect noisy
measurements.

[Link] Regularizing of Inverse Problem 2 / 22

Concepts Under Inverse Problem

[Link] axial tomography

Computational Axial Tomography (CT) uses X-rays to create images
of objects. Regularization techniques are applied in CT to improve
image accuracy by overcoming challenges like noise and limited data,
ensuring more reliable reconstructions.
[Link]
Deconvolution is a method used in inverse problems to remove
blurring effects from observed signals or images, aiming to recover
clearer and more accurate details of the original data.

[Link] Regularizing of Inverse Problem 3 / 22

[Link]
When travelling in a boat or plane, it is useful to have an idea of the
current location in close to real time. This is often done by making a
variety of measurements, for example by usingbearings to landmarks,
stellar or satellite positions, and also by considering one’s previous
position and using information such as records of speed and heading.
[Link] analysis
Image analysis in an inverse problem means figuring out an object’s
properties from its images. For instance, creating a 3D structure
from 2D medical scans or restoring an image by removing distortions.
It involves solving mathematical models to uncover the object’s
characteristics from the available image data.

[Link] Regularizing of Inverse Problem 4 / 22

[Link]
Inverse problems have always played an important role in geophysics
as the interior of the Earth is not directly observable yet the surface
manifestation of waves that propagate through its interior are
measurable. Using the measurements of seismic waves to determine
the location of an earthquake’s epicentre, or the density of the rock
through which the waves propagate, are typical of inverse problems in
which wave propagation is used to probe an object.

[Link] Regularizing of Inverse Problem 5 / 22

Ill-posed problems and ill-conditioning

Ill-posed problems and ill-conditioning are key challenges

encountered in the context of regularizing inverse problems:
Ill-posed problems:-These refer to problems that lack unique
solutions, stability, or sensitivity to data perturbations. Inverse
problems are often ill-posed because they might have infinitely many
solutions or no solution at all. When the input data is noisy or
incomplete, the inverse problem becomes more challenging as it
might not have a unique and stable solution without additional
constraints or regularization.
Ill-conditioning:- This describes a situation where small changes or
perturbations in the input data can lead to significant changes in the
output or solution. Ill-conditioned problems are
[Link] Regularizing of Inverse Problem 6 / 22
sensitive to slight variations in the input data, making them prone to
numerical instabilities and inaccuracies in the computed solutions.
these methods help in finding more meaningful, stable, and reliable
solutions despite the ill-posed or ill-conditioned nature of the
problem. Regularization helps to mitigate the effects of noise and
uncertainties, preventing overfitting and improving the quality of the
reconstructed solutions in inverse problems.
Ill-posed problems to well posed
An ill posed problem is defined as a problem that is not
well-posed,there are three criteria for classifying a macth problem as
well-posed.A problem is said to be well-posed if:

The problem has solution,

The solution is unique
The solution depends continuously on the data

[Link] Regularizing of Inverse Problem 7 / 22

 Let’s consider the problem of dividing a number by zero. Dividing
any number by zero is an example of an ill-posed problem in
mathematics because it does not have a unique [Link], let’s
transform this ill-posed problem into a well-posed one by slightly
modifying the problem. Instead of dividing a number by zero directly,
let’s find the limit of a number divided by a very small value that
approaches zero.
- Initially: x0 (ill-posed because dividing by zero is undefined) -
Transforming: limy →0+ yx (well-posed, as it approaches zero from a
positive value)
By using limits and allowing the divisor to approach zero from a
positive value, we transform the previously ill-posed problem of
dividing by zero into a well-posed problem, enabling a defined and
meaningful solution as the divisor gets infinitesimally close to zero
without actually being zero.

[Link] Regularizing of Inverse Problem 8 / 22

Applications

[Link] Problem in Medical Imaging

What is an inverse problem in medical imaging?
The inverse problem in medical imaging refers to imagine trying to
figure out what something looks like inside your body based on the
pictures taken by machines like X-rays or [Link] involves determining
the characteristics or structure of the object from the measurements
taken by the imaging device,It’s a bit like solving a puzzle without
having all the pieces, where you have to guess what the object or
tissue looks like based on the information you have from those
pictures. It’s a tricky task because you need smart computer
programs to help put all the information together accurately to create
an image of what’s going on inside you.

[Link] Regularizing of Inverse Problem 9 / 22

X-Ray CT
In X-ray CT (Computed Tomography), the inverse problem involves
reconstructing the internal structure of an object based on X-ray
measurements taken from multiple angles. Mathematically, it’s about
determining the distribution of X-ray attenuation coefficients within
the imaged object.
Here’s a simplified explanation:
Forward Problem:X-rays are sent through the body, and
detectors measure the intensity of transmitted X-rays at different
angles.
Inverse Problem:The challenge is to reconstruct the original
distribution of X-ray attenuation within the body from these
measurements. This is often solved using mathematical
algorithms, such as iterative reconstruction methods.

[Link] Regularizing of Inverse Problem 10 / 22

Effect of Noise:Noise in the measurements or reconstruction
process can introduce artifacts or inaccuracies in the final image.
Regularization techniques and advanced algorithms are used to
mitigate these effects and improve the accuracy of the
reconstructed CT images.

[Link] Regularizing of Inverse Problem 11 / 22

Positron emission tomography
In principle, the model for PET is the same as that for X-Ray
CT, however, there is so much noise in the measurements.
The images have much less resolution.
They look good, but are very small.

Figure: PET image

[Link] Regularizing of Inverse Problem 12 / 22

[Link] Problems In Remote Sensing
Satellite remote sensing
Satellite remote sensing has become an important way to monitor
and study our environment.
Land, vegetation, oceans, snow, ice, atmosphere,
Global observations from pole to pole.
Continous observations,
monitoring, forecasting, support in emergency situations.
support for decision making .

[Link] Regularizing of Inverse Problem 13 / 22

Satellite remote sensing and inverse problems

Satellite remote sensing involves collecting Earth data from space.

Inverse problems in this field occurs data to extract specific
information, like removing atmospheric interference, classifying land
cover, estimating temperature, or mapping elevation. Mathematical
models and algorithms are used to solve these problems and retrieve
meaningful insights about the Earth’s surface or atmosphere from
satellite observations.

[Link] Regularizing of Inverse Problem 13 / 22

Measurement error, noise

By repeating the measuremenets we get different answers.

for example,let us consider game of casino and dice. The
resultant values changes and gives a different results each and
every time it’s performed.

[Link] Regularizing of Inverse Problem 14 / 22

[Link] Problems In Greenhouse Gas Modelling
The inverse problem in greenhouse gas refers to the process of
determining the sources or emissions of these gases based on
measurements collected from the atmosphere. Instead of measuring
emissions directly, scientists use atmospheric data to infer the origin
and quantity of greenhouse gases released from various sources, like
industrial activities, transportation, or natural processes. It involves
complex mathematical models and atmospheric measurements to
estimate and track the sources of these gases.
In greenhouse gas modeling, inverse problems help figure out where
emissions come from based on the air’s gas levels. There are different
smart methods for this:
Dispersion Models: Simulate how gases move in the air to
guess where emissions are coming from.

[Link] Regularizing of Inverse Problem 14 / 22

Atmospheric Inversion: Use data from different sources to
estimate emissions, improving the guesses over time.
Satellite Sensing: Satellites measure gases in the air, helping
to find emission sources from above.
Optical Sensing and Isotopic Analysis: Use ground-based
sensors to study gas concentrations and apply methods to
distinguish different emission sources.
Machine Learning:Teach computers to recognize patterns in
data for better emission estimates.
All these methods help us understand where greenhouse gases
are coming from, which is crucial for creating plans to reduce
emissions and fight climate change.

[Link] Regularizing of Inverse Problem 15 / 22

[Link] Problems In Image Processing
In image processing, the concept of an ”inverse problem” typically
refers to the challenge of recovering unknown parameters or original
data from observed or processed data. It’s a complex issue with
various applications like image reconstruction, deblurring, or
denoising. Specific techniques depend on the nature of the problem
and the type of imaging involved.
Deconvolution: restore blurred and noisy image.
- Applications: most imaging applications.
Inverse halftoning: .restores the quality of image
- Applications: improves print quality.

[Link] Regularizing of Inverse Problem 16 / 22

Deconvolution

Problem y = x ? h + n; given y,h, find x

Applications: most imaging application (seismic, medical,
satellite)

[Link] Regularizing of Inverse Problem 17 / 22

Halftoning and inverse halftoning

Halftoning (HT):Halftoning is a method used in printing or

displaying images with dots to simulate continuous
tones or colors.
Inverse halftoning (IHT): Halftoning is a method used in
printing or displaying images with dots to simulate continuous
tones or colors.

[Link] Regularizing of Inverse Problem 18 / 22

[Link] Problems in Breast Tumour Detection
Inverse problems in breast tumor detection often involve
reconstructing the internal properties of breast tissues from external
measurements. Potential solutions include:
Tomographic Imaging Techniques: Methods like X-ray
mammography, MRI, or ultrasound can be used to capture
internal images, aiding in tumor detection.
Computational Algorithms:Computational algorithms assist
by analyzing imaging data to detect, characterize, and quantify
breast tumors accurately, aiding in precise diagnosis and
treatment planning.
Machine Learning:These algorithms learn from vast datasets
of images, assisting in accurate tumor identification,
classification, and providing treatment strategies.

[Link] Regularizing of Inverse Problem 19 / 22

 Electrical Impedance Tomography (EIT): Utilize electrical
measurements to reconstruct tissue conductivity, potentially
revealing anomalies associated with tumors.
Optical Imaging: Employ techniques like diffuse optical
tomography to use light propagation through tissues for imaging
and tumor detection.
Hybrid Imaging: Combine multiple imaging modalities for
improved accuracy, such as PET-MRI.

[Link] Regularizing of Inverse Problem 20 / 22

Conclusion

This project underscores the pivotal role of regularization

techniques in addressing complex inverse problems across diverse
domains. The application proves instrumental in stabilizing ill-posed
problems and enhancing the accuracy of [Link] the context of
greenhouse gas modeling,contributing to effective climate change,
medical imaging and image processing of regularization methods
improves image reconstruction, remote sensing emphasizes the
importance of stabilizing inverse [Link] technique
helps in early detection and treatment planning for breast tumor.
This project highlights that the judicious application of regularization
is fundamental in unraveling complex inverse problems.

[Link] Regularizing of Inverse Problem 21 / 22

BIBLIOGRAPHY

Albert Tarantola, Inverse Problem Theory, and Methods for

Model Parameter Estimation. SIAM, 2005.
W. Alt, Linear Functional Analysis, An Application-Oriented
Introduction, Universitext, Springer, London, (2016).
R. Andreev, P. Elbau, M. V. de Hoop, L. Qiu , O. Scherzer,
Generalized convergence rates results for linear inverse problems
in Hilbert spaces, Numer. Funct. Anal. Optim. 36(5),pp 549–566,
(2015).
A. B. Bakushinski, Remarks on choosing a regularization
parameter using the quasioptimality and ratio criterion, USSR
Comput. Math. Math. Phys. 24(4),pp 181–182, (1985).

[Link] Regularizing of Inverse Problem 22 / 22

 Thank You

[Link] Regularizing of Inverse Problem 23 / 22