SPECKLE REDUCTION, SEGMENTATION AND REGISTRATION IN MEDICAL IMAGES : A COMPARATIVE STUDY

ByAbhinav Kumar KushwahaAlok Ranjan

SupervisorDr. Rajeev Srivastava

ABSTRACT Speckle noise is a ubiquitous artifact that limits the

interpretation of optical coherence tomography images and ultrasound images. Here we apply various speckle-reduction digital filters to OCT and US images and compare their performances. Our results indicate that adaptive filters, enhanced Lee, Frost and Wiener filters can significantly reduce speckle and increase the signal-to-noise ratio, while preserving strong edges.

In computer vision, segmentation refers to the process of partitioning a digital image into multiple segments (sets of pixels, also known as super pixels). The goal of segmentation is to simplify and/or change the representation of an image into something that is more meaningful and easier to analyze.

Image registration is the process of overlaying images of the same scene taken at different times, from different viewpoints, and/or by different sensors. The registration geometrically aligns two images (the reference and sourced image).

ABSTRACT…. Here, we present a tool which can perform speckle

reduction, segmentation and registration all using several methods and then compare the results obtained.

KEYWORDS: Speckle Reduction, Medical Image, Image Segmentation, Registration, Ultrasound Image, Image Processing, SRAD filter, Lee Filter, Frost Filter, PDE based filter, Locally affine, FFT based algorithm, Tool design, Performance Comparison.

SPECKLE PATTERN Random intensity pattern produced by the

interference of a set of wave fronts This phenomena has been investigated by

scientists since the time of Newton An illuminated surface acts as a source of

secondary spherical waves At any point in this scattered light field, the

light is made up of waves of different path lengths and the resultant wave varies randomly

If light of low coherence is used, a speckle pattern will not be normally observed

Speckle patterns can be observed in polychromatic light

SPECKLE REDUCTION

Original Holographic image (left) and Speckle reduced image (right)

SPECKLE REDUCTION

Original Ultrasound image (left) and Speckle reduced image (right)

SEGMENTATION Segmentation refers to the process of partitioning

a digital image into multiple segments. More precisely, image segmentation is the

process of assigning a label to every pixel in an image such that pixels with the same label share certain visual characteristics.

Each of the pixels in a region is similar with respect to some characteristic or computed property, such as color, intensity, or texture.

The result of image segmentation is a set of segments that collectively cover the entire image, or a set of contours extracted from the image.

EXAMPLE OF TEXTURE BASED SEGMENTATION

REGISTRATION Fusion, Superimposition and Matching A process of finding a transformation that

aligns one image to another Goal of registration is to find a

correspondence function or mapping M(.) that takes each spatial co-ordinate X(s) for source image and returns a co-ordinate X(t) for the target image

Used in comparing images to find tumor growth, target tracking in defense, face/thumb recognition in security

IMAGE REGISTRATION

METHODS USED FOR IMAGE REGISTRATION Locally Affine Based Projective Based Fast Fourier Transform Based MIRT tool

LOCALLY AFFINE BASED In the first stage, the transformation is

formulated as being purely affine. In the second stage, this purely or affine

geometric transform is extended to also implicitly account for contrast and brightness modulations.

Finally, in third stage, a smoothness constraint is imposed on all locally estimated geometric and intensity parameters.

PROJECTIVE BASED First we calculate vertical and horizontal

derivatives between the two images. From these derivatives, we estimate an

approximate model of the projective parameters such as bilinear.

Calculate the new coordinates from the approximate model

These old and new coordinates now completely determine the projective parameters in the exact model

These new parameters are now applied to one of the images and iterate till negligible difference

FAST FOURIER TRANSFORM BASED This method relies on the translation

property of the Fourier transform referred to as Fourier Shift theorem.

By taking inverse Fourier Transform of the representation in the frequency domain, we have a function which is approximately zero everywhere except at the displacement that is needed to optimally register the two images.

This method shows excellent robustness against random noise.

MIRT MIRT is a Matlab software package for 2D

non-rigid image registration. The geometric transformation is based on

cubic B-splines. The optimization is based on Euler Method

(gradient-based).

RESULTS

Locally Affine Based

RESULTS Projective Based

RESULTS FFT based

RESULTS MIRT

RESULTS Target image is obtained by adding

brightness to the source image and comparison has been done between locally affine based and MIRT.

RESULTS Target image has been obtained by varying

the contrast in the source image and comparison has been done between locally affine based and MIRT.

RESULTS Target image is obtained by scaling the

source image by varying factor and the behavior of FFT based algorithm has been observed.

RESULTS Target image is obtained by rotating the

source image by varying the angle and the behavior of projective based algorithm has been observed.

CONCLUSION Performance of MIRT is better when

variations are performed on brightness and contrast of the target image whereas FFT and projective based approach performs very poor.

FFT based approach has better results when scaling has been performed on target image.

When target image is rotated by different angles then projective based approach has best results than other three approaches.

REFERENCES[1] Society of Nuclear Medicine[2] Dorland's Medical Dictionary for Health Consumers, 2007 by Saunders; Saunders Comprehensive Veterinary Dictionary, 3 ed. 2007; McGraw-Hill Concise Dictionary of Modern Medicine, 2002 by The McGraw-Hill Companies[3] Dhawan P, A. (2003). Medical Imaging Analysis. Hoboken, NJ: Wiley-Interscience Publication[4] Linda G. Shapiro and George C. Stockman (2001): “Computer Vision”, pp 279-325, New Jersey, Prentice-Hall[5] Ron Ohlander, Keith Price, and D. Raj Reddy (1978): “Picture Segmentation Using a Recursive Region Splitting Method”, Computer Graphics and Image Processing, volume 8, pp 313-333[6] S. Osher and N. Paragios. Geometric Level Set Methods in Imaging Vision and Graphics, Springer Verlag, ISBN 0387954880, 2003.

REFERENCES…..[7] Jianbo Shi and Jitendra Malik (2000): "Normalized Cuts and Image Segmentation", IEEE Transactions on pattern analysis and machine intelligence, pp 888-905, Vol. 22, No. 8[8] Leo Grady (2006): "Random Walks for Image Segmentation", IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1768-1783, Vol. 28, No. 11[9] Z. Wu and R. Leahy (1993): "An optimal graph theoretic approach to data clustering: Theory and its application to image segmentation", IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1101-1113, Vol. 15, No. 11[10] Leo Grady and Eric L. Schwartz (2006): "Isoperimetric Graph Partitioning for Image Segmentation", IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 469-475, Vol. 28, No. 3[11] C. T. Zahn (1971): "Graph-theoretical methods for detecting and describing gestalt clusters", IEEE Transactions on Computers, pp. 68-86, Vol. 20, No. 1[12] Efficient image segmentation using partial differential equations and morphology, Joachim Weickert[13] Pde-Based Modeling of Image Segmentation using Volumic Flooding by Anastasia Sofou and Petros Maragos[14] Comparison of PDE based and other techniques for speckle reduction from digitally reconstructed holographic images by Dr. Rajeev Srivastav, JRP Gupta and Harish Parthasarthy

THANKS