image fusion using evolutionary algorithm (ga)

IMAGE FUSION USING EVOLUTIONARY ALGORITHM (GA)

V Jyothi 1), B Rajesh Kumar 1), P Krishna Rao 2), D V Rama Koti Reddy 2)

1) GITAM University, Visakhapatnam, AP, India, [email protected]

2) Andhra University, Visakhapatnam, AP, India, [email protected]

Abstract: Image fusion is the process of combining

images taken from different sources to obtain better

situational awareness. In fusing source images the

objective is to combine the most relevant information

from source images into composite image. Genetic

algorithm is used for solving optimization problems.

Genetic algorithm can be employed to image fusion where

some kind of parameter optimization is required.

In this paper we proposed genetic algorithm based

schemes for image fusion and proved that these schemes

perform better than the conventional methods through

comparison of parameters namely image quality index,

mutual information, root mean square error and peak

signal to noise ratio.

Keywords: Genetic Algorithm, Image quality Index,

Mutual Information.

1. INTRODUCTION For remotely sensed images, some have

good spectral information and the others have

geometric resolution, how to integrate these two

kinds of images into one image is a very

interesting thing in Image processing, which is

also called image fusion.

Image fusion is emerging as a vital

technology in many military, surveillance and

medical applications. It is a sub area of the more

general topic of data fusion, dealing with image

and video data. The ability to combine

complementary information from a range of

distributed sensors with different modalities can

be used to provide enhanced performance for

visualization, detection or classification tasks.

Multi-sensor data often present complementary

information about the scene or object of interest,

and thus image fusion provides an effective

method for comparison and analysis of such

data. There are several benefits of multi-sensor

image fusion: wider spatial and temporal

coverage, extended range of operation,

decreased uncertainty, improved reliability and

increased robustness of the system performance.

In several application scenarios, image

fusion is only an introductory stage to another

task, e.g. human monitoring. Therefore, the

performance of the fusion algorithm must be

measured in terms of improvement in the

following tasks. For example, in classification

systems, the common evaluation measure is the

number of the correct classifications. This

system

evaluation requires that the”true” correct

classifications are known. However, in

experimental setups the ground-truth data might

not be available. In many applications the

human perception of the fused image is of

fundamental importance and as a result the

fusion results are mostly evaluated by subjective

criteria. Objective image fusion performance

evaluation is a tedious task due to different

application requirements and the lack of a

clearly defined ground-truth. Various fusion

algorithms presented in this project. Several

objective performance measures for image

fusion have been proposed where the

knowledge of ground-truth is not assumed.

There are many Image Fusion

techniques based on signal, pixel, feature and

symbol level fusion. In many situations, a

single image cannot depict the scene properly.

In these cases, scene is captured through more

than one sensors, but human and machine

processing is better suited with a single image,

so therefore we need to fuse the images

obtained from different sensors to obtain a

single composite image which contains relevant

information of source images.

2. GENETIC ALGORITHM A variety of algorithms have been

evolved from nature. Genetic algorithm is one

of the simplest and most popular evolutionary

VJyothi,B.Rajesh Kumar,P.Krishna Rao,D.V.Rama Koti Reddy, Int. J. Comp. Tech. Appl., Vol 2 (2), 322-326

322

ISSN:2229-6093

algorithms. Genetic Algorithms (here onwards

called as GA) are based on natural selection

discovered by Charles Darwin. GA makes use

of the simplest representation, reproduction and

diversity mechanism. Optimization with GA is

performed through natural exchange of genetic

material between parents. Offspring’s are

formed from parent genes. Fitness of offspring’s

is evaluated. The fittest individuals are allowed

to breed only.

GA's are being used in different

applications such as function Optimization,

System Identification and Control, Image

Processing, Parameter Optimization of

Controllers, Multi-Objective Optimization, etc.

Algorithm

• Choose initial population

• Evaluate the fitness of each individual in

population

• Repeat

• Select best-ranking individuals to

reproduce a new population

• Breed new generation through crossover

and mutation to give birth to offspring

• Evaluate the individual fitness of the

offspring

• Replace worst ranked part of population

with offspring

• Until some termination condition is met

3. IMAGE FUSION TECHNIQUES

Pixel level Average method

This technique is a basic and straight forward

technique and fusion could be achieved by simple

averaging corresponding pixels in each input image

as:

Pixel level Weighted average method

We add some weights to the individual images and

perform the averaging technique as follows:

where W1 and W2 are the weights.

Pixel level weighted average method using GA

In this method the weights are estimated using the

GA and a new optimized image is obtained from the

average method using the optimized weights.

Where GA(W1) is the optimized value of weight

W1 and GA(W2) is the optimized value of weight

W2.

1. DWT based image fusion

In wavelet image fusion scheme, the source images

I1(a, b) and I2(a, b) are decomposed into

approximation and detailed coefficients at required

level using DWT. The approximation and detailed

coefficients of both images are combined using

fusion rule f. The fused image could be obtained by

taking the inverse discrete wavelet transform

(IDWT) as:

The fusion rule used is simply averages the

approximation coefficients and picks the detailed

coefficient in each sub band with the largest

magnitude.

2. Weighted average DWT based image fusion

In this method additional weights are selected along

with the DWT of the images. The fused image can

be obtained by taking the inverse discrete wavelet

transform (IDWT) as:

3. Weighted average DWT based image fusion using

GA

In this method additional weights are estimated

using GA along with the DWT of the images. The

fused image can be obtained by taking the inverse

discrete wavelet transform (IDWT) as:

4. EVALUATION CRITERIA

Objective image quality measures play an important

role in various image processing applications. There

are different types of object quality or distortion

assessment approaches. The fused images are

evaluated, taking the following parameters into

consideration.

Root Mean Square error (RMSE)

The root mean square error (RMSE) between each

unsharpened MS band and corresponding sharpened

band can also be computed as a measure of spectral

fidelity. It measures the amount of change per pixel

due to the processing.


323

ISSN:2229-6093

The RMSE between a reference image R and the

fused image F is given by

There are different approaches to construct reference

image using input images. In our experiments, we

used the following procedure to compute RMSE.

First, RMSE value El is computed between source

image A and fused image F.

Similarly E2 is computed as RMSE between source

image B and fused image F.

Then the overall RMSE value is obtained by taking

the average of E1 and E2.

Smaller RMSE value indicates good fusion quality.

Peak Signal to Noise Ratio

PSNR can be calculated by using the formula

Where MSE is the mean square error and L is the

number of gray levels in the image.

Image Quality Index

IQI measures the similarity between two

images (I1 & I2) and its value ranges from -1 to 1.

IQI is equal to 1 if both images are identical. IQI

measure is given by

Where x and y denote the mean values of

images I1 and I2 and , , and denotes the

variance of I1 , I2 and covariance of I1 and I2.

Mutual Information

Mutual Information (MI) measures the

degree of dependence of two images. Its value is

zero when I1 and I2 are independent of each other.

MI between two source images I1 and I2 and fused

image F is given by

and PA(a) ,PB(b) and PF(f) are histograms of images

A, B and F,PFA(f,a) and PFB(f,b) are the joint

histograms of F and A, and F and B respectively.

Higher MI value indicates good fusion results.

RESULTS

We have taken a medical image to evaluate the

results by Averaging method and Satellite

images for evaluating the images by DWT

method.

Input image 1

Figure : CT image

Input Image 2

Figure 2 : MR image

Fused Image by Averaging method


324

ISSN:2229-6093

Image Fused by GA Average Method

Input Image 1

Input Image 2

Fused Image by DWT Method

Fused Image by GA – DWT


325

ISSN:2229-6093

Performance Comparison of Proposed

Schemes

METHOD IQI MI RMSE PSNR

GA_AVG 0.9851 1.1293 12.3288 26.3124

GA_DWT 0.9468 1.0042 20.6849 21.8177

CONCLUSION

They are many ways of fusing images. We have

compared the regular image fusion techniques

with the Genetic Algorithm based techniques. It

can be seen from the above table and the image

results that the GA based techniques are having

much better results when compared with the

conventional techniques.

Two Genetic Algorithm based image fusion

algorithms are introduced and their objective

and subjective comparison with other classical

techniques is carried out. It is concluded from

experimental results that GA based image

fusion schemes perform better than existing

schemes.

6. REFERENCES [1] Aqeel Mumtaz*, Abdul Majid, Adeel Mumtaz

“Genetic Algorithm and its applications to

Image Processing”. 2008 International

Conference on Emerging Technologies, IEEE-

ICET 2008

[2] A.Haq Nishat, "Multi-Sensor Image Fusion and

Image Colorization for Better Situation

Assessment", Master Thesis, GIKI Pakistan,

Dec 2005. (PI)

[3] A M Khan, A Khan,” Fusion of Visible and

Thermal Images using Support Vector

MachinesT. Scientist. Title of the paper.

Proceedings of the Workshop “Intelligent Data

Acquisition and Advanced Computing Systems:

Technology and Applications

(IDAACS’2001)”, Ternopil, Ukraine 1-4 July

2001, pp.123-127.

[4] G. Piella, “A general framework for multiresolution image fusion: from pixels to regions,” Information Fusion, vol. 4, pp.


326

ISSN:2229-6093

image fusion using evolutionary algorithm (ga)

Documents