A nonlinear edge –preserving smoothing filter for edge detection on color and gray satellite images

A nonlinear filter for smoothing color and gray images corrupted by Gaussian noise is presented in this paper. The proposed filter designed to reduce the noise in the R,G, and B bands of the color images and preserving the edges. This filter applied in order to prepare images for further processing such as edge detection and image segmentation. The results of computer simulations show that the proposed filter gave satisfactory results when compared with the results of conventional filters such as Gaussian low pass filter and median filter by using Cross Correlation Coefficient (ccc) criteria. Edge detection, nonlinear filters, Gaussian noise, Sobel operator, color image.


Introduction
Visual information processing is increasingly becoming widespread as multimedia becomes common in everyday life. With the expanding use of color images in multimedia applications and the proliferation of color capturing and display units, the interest in color imaging is rapidly growing. During image formation, acquisition, storage and transmission many types of distortions limit the quality of digital images. Edge detection plays a very important role in the realization of a complete image understanding system. Indeed, high-level processing tasks such as image segmentation and object recognition directly depend on the quality of the edge detection procedure. It is well known, however, that the generation of an accurate edge map becomes a very critical issue when the images are corrupted by noise. In this respect, noises having Gaussian-like distribution are very often encountered during image acquisition. Methods that represent extensions of monochrome edge detector typically process the threecolor channels independently and combine the resulting edge maps. As an example, the application of the well-known Sobel technique to the three image channels can be considered at a great aspect. An edge point in the color image could be estimated by evaluating the maximum of the gradient Components or their vector sum [1].
Edge detection in colored images has not gained much attention as in grayscale images. 10% more edges can be detected in colored images due to change in color. In colored images, color vector is assigned to each pixel because color component comprises of three color values [2].
Different methods to color edge-preserving and edge detection in presence of kind of noise have been proposed in the scientific literature.
C. Tomasiand R. Manduchi (1998) [3] has introduced the concept of bilateral filtering for edgepreserving smoothing. Bilateral filtering smooth's images while preserving edges, by means of a nonlinear combination of nearby image values. Carsten Garnica, Frank Boochs, and Marek Twardchlib(2000) [4] are presented new approach for elimination noise within digital images this an essential step in applications like edge detection or image segmentation, the algorithm is based on the edge-preserving Maximum Homogeneity Neighbour Filter. A new technique for edge detection in color images corrupted by Gaussian noise is presented by Fabrizio Russo, and Annarita Lazzari (2005) [5], the proposed approach adopts a multipass processing architecture that aims at reducing noise before extracting the image edges. Meng-Husiun Tsai, Yung-Kuan Chan, Zhe-Zheng Lin, Shys-Fan Yang-Mao, and Po-Chi Huang(2011) [6]present a cytoplast and nucleus contour (CNC) detector to sever the nucleus and cytoplast from a cervical smear image after smoothing image. They adopted median filter to sweep off noise. Yang Ou, and Dai Guang Zhi (2011) [1] suggested a nonlinear filtering to reduce Gaussian noise based on multichannel data fusion before edge detection. This method could smooth noise and preserve the details of edge without any a priori knowledge about the distribution of noise. Payun S. Tembhurne and Prof. V.K. Shandilya (2012) [7], proposed color image processing pipeline provide good reference, which can be implemented using multistage bilateral filters that integrates two types of noise filtering to identify the edges and smooth out the noises. Upneet Kaur(2013) [8], present Gaussian filter to smoothing color image before edge detection.

Smoothing filters types 1. Lowpass filters
The most commonly used smoothing algorithms are the n*n-box filter and the Gaussian Kernel Smoothing (Binomial Filter). Here, a square-sized convolution kernel is applied to each pixel. With the parameter "box size" (n) respectively "sigma", the degree of smoothing can be controlled. This smoothing strategy results in very nicely smoothed homogeneous areas, however possessing the disadvantage to blur gray value edges [4]. • Conditional Averaging Filter These nonlinear algorithms are calculating the filtered gray value in dependence of the content of a defined neighbourhood. Each edge-preserving filter has its own specific algorithm, but they all have in common, that the effect of this smoothing strategy is to preserve edges [4].

3-The proposed filter
The proposed method of filtering based on sliding neighbor operation where the new value of an output pixel is equal to the difference between the maximum value and the minimum value of the selected pixels under mask divided by the mean of value of pixels under the same mask as the following: 1

Edge detection using Sobel operator
The Sobel operator is the magnitude of the gradient computed by [9]: ( 2 ) and can be implemented using convolution masks: Note that this operator is placed on an emphasing pixel that is closer to the center of the mask. The Sobel operator is one of the most commonly used edge detection [9].

Cross correlation coefficient (ccc)
The Cross Correlation Coefficient (ccc) which is used to measure the quality of an image. If we have two images, the original image f(x,y) and the processed image g(x,y), the (ccc) is given by [10]: Where 0 1 , if ccc=1 means a perfect correlation , but when 0 means totally uncorrelated image [10].

Experimental work and results
Edge is high frequency also noise is high frequency it is particularly important to reduce the effects of noise before edge detection to improve the results of edge detection, so the gradient preserves the noise much more than structure to avoid this focused by smoothing image first which removes noise much more than it removes structure this mean preserving the desired details of image.
For testing the performance of the proposed filter we have focused on the filtering of images using the linear and nonlinear filtering techniques where linear filtering is done using the Gaussian while the nonlinear filtering is performed using a median filter then compared them with the proposed filter.
We implement the proposed, Gaussian and median filters on both color and gray images for the sake of comparison between the results of these filters.
In the case of color image the procedures of processing by using Matlab programming's are: 1-The original image which is taken "24-bits RGB Color Satellite image "(24-bits i.e. 8bits for each of the three color bands) with size (256 by 256) pixels and 1 meter resolution (university of Baghdad zone)as shown in Fig.1-a, where R( Fig. (3a).   1: (a)24-bit RGB color Satellite image (b)Red-band, (c)Green-band, and (d)    While in the case of the gray image the procedure of the processing are: 1-The original image which is taken with size (256 by 256) pixels (university of Baghdad zone) as shown in Fig.(5-a). 2-A noisy images, were generated by adding zero-mean Gaussian noise starting with standard deviation σ=0.005 up to σ=0.01 to examine the filter's ability to reduce noise and preserving edges as shown in Fig.5 Fig.7.