Image Segmentation based on Fuzzy Genetic Algorithm

— Image segmentation is one of the most important tasks to extract information in image processing. To satisfy increasing requirement of image segmentation, a variety of segmentation methods have been developed over the past several years. Fuzzy c-means (FCM) is unsupervised segmentation technique that has been successfully applied to future analysis, clustering, and classification but the FCM and its derivative algorithms suffer from various noises in the images. In this paper a new Fuzzy Genetic (FG) algorithm is introduce. This method describes Chromosome representation, Population Initialization, Fitness function, Genetic Operators, domain of the parameters collectively for image segmentation which produced better result comparatively FCM and its derivative methods.


A. Noise reduction
In this paper, it distinguishes noise with respect to local neighborhood of each pixel. In addition, for detection of noise, consider the neighboring pixels. The main objective of fast image segmentation method, it needs to make use of fast noise reduction method.

i) Detection of the noise pixels
To detect a noisy pixel with the help of neighboring pixel, it follows the method proposed by efficient impulse noise reduction and it defines the neighborhood of a pixel as shown in figure 1. Distance of central pixel is calculated using the following formula 1 5 tan with NW z (1) Dis ce z   The neighbor pixel was obtained based on distance of the corresponding pixels. These distances were sorted in an ascending order, and 5 neighbors with longest distances are selected. If the average of these 5 values was greater than a certain threshold, it means that at least five of the neighboring pixel has been very different so that the pixel is not considered as an edge position, and it is probably a noise pixel.
ii) Correction of the noisy pixels Neighboring pixels was corrected by identifying pixel as a noisy. The basic reason is, it uses neighboring pixel lies in the fact that a pixel is expected to be similar to its surrounding pixel. To do this, for selecting the pixel where total distance of its neighbor pixels with neighboring pixels of noisy pixel is less; this means that Euclidean distance between each neighboring pixel selected from the neighboring pixel in the same position of noisy pixel is calculated. Pixels to produce the lowest value mean that their neighbors are more like together. Two pixels with similar neighbor values are expected to be replaced y the noisy pixel.

B. FCM
FCM is an iterative clustering that produces an optimal cluster partition by minimizing the weighted group sum of squared objective function   .
is the data set in the  -dimensional vector space, is the number of data items,  is the number of clusters with     2 , ij  is the degree of membership of i  in the j th cluster,  is the weighting exponent on each fuzzy membership, j  is the prototype of the center of cluster j,

B. Population initialization:
To construct the initial population using the method described earlier. To do this we create arbitrary number of randomly initialized chromosomes. In GAs, the initial population consists of random strings. However, random binary strings, each of the length pXq (q bits for each of the parameters) can be considered as chromosomes or individuals of the initial population. C. Fitness function: The reproduction is the process in which individual strings are copied according to their object function values. Objective function values are denoted by F and is called as Fitness function. In an enhanced image the fitness is measured by the sum of intensities of edges. The fitness function is evaluated by using two steps. First step, the pixels data set is clustered according to the centers encoded in the chromosomes under consideration. Let the each intensity value x i , i=1, 2, 3… mxn is assigned to cluster with center z j , j=1, 2… K.

D. Genetic Operators:
To produce the solution at each generation, Genetic algorithm uses the principle of selection. By using cross over and mutation Matting of parents are represented. i) Selection: For selecting the next generation individuals, the selection operation is used. Reproduction is also called selection operator. In Genetic algorithm, Roulette wheel selection is most widely used technique. If the fitness of individual cluster C i in the population, then its probability is calculated as Where N is the number of individuals in the population. ii) Crossover:To recombine the information Crossover operator is used. Crossover is a genetic parameter and it used to combine two chromosomes called as parents to produce new chromosome called as child (also called as offspring) chromosome. The child chromosome is also called as offspring. The result of Crossover provides a new chromosome may be better than original chromosomes. Child chromosome will have some properties from one parent and other properties from other parent. Suppose parent1 is 10110100 and parent2 is 10100011 and after performing the crossover the result which contains some part of parent1 and other from parent2. iii) Mutation: After crossover, mutation is performed and it depends on the encoding and crossover. It maintains genetic diversity from one generation of a population of chromosomes to the next. Offspring are changed randomly by Mutation operation. For a binary encoding we can switch few randomly chosen bits and it changes bits from 0 to 1 or 1 to 0. ( 1 2 ) P Q Jac P Q P Q

  
Where P and Q are both non -empty, we define jac (P, Q) =1 and 1 ) ,

V. RESULTS AND ANALYSIS
To demonstrate the effectiveness and robustness we conduct few experiments. The proposed method Fuzzy Genetic Algorithm is tested and compare the existing FCM method 100 images data set.Binary performance is evaluated by using the Jaccard Index(JAC), mean(  ) and standard deviation( ). In this paper Fuzzy Genetic algorithm is used for image enhancement. The result analysis of this paper is divided into two methods subjective evaluation and experimental evaluation. In this section subjective evaluation of Fuzzy Gentic method. We compare the results of Fuzzy Genetic algorithm method with FCM method. Our porposed metod provides the better results for Lena, bird, Crow, two birds, three birds,elephant, things and wheel.The figure shows the comparision of existing and proposed methods and it also displays the original image histograms. In img7 provids the better solution for uneven illuminatios.    VI. CONCLUSION In this paper new method is proposed by using Fuzzy Genetic algorithm. FCM and its derivative methods suffer from noise in the images. To overcome this drawback the proposed technique is used. It very useful for object reorganization even the image contains noisy, shadows and low contrasted images. The proposed method effectively removes shadows in image, noise in the image and identified objects even for low contrasted images. We evaluated the performance of the proposed method 3 different performance measures like mean, standard deviation and Jaccard index. This method gives better results.