4 Proposed method:

4.1 Bi-Dimensional Empirical Mode Decomposition: The Empirical Mode Decomposition (EMD) was a technique for analyzing nonlinear and nonstationary signals. It serves as an alternative to methods such as wavelet analysis and short-time Fourier transform. It decomposes any complicated signal into a finite and often small number of Bi-dimensional Intrinsic Mode Functions (IMF). The decomposition method in EMD is called Shifting Process. The shifting process of the two- dimensional signal such as image can be adapted from the one-dimensional signal. In this project, we introduce the concept of Multi sensor image fusion technique for impulse noise reduction in digital images using BiDimensional Empirical Mode Decomposition. The images captured by two different sensors undergo filtering using two different vector median filtering algorithms. The filtered images are decomposed into a small number of Bi-dimensional Intrinsic Mode Functions (IMF). Different fusion rules are applied to different IMFs obtained from BEMD. Finally all fused components are combined to get the noise free image. 4.2 Intrincic mode decomposition(IMF) An IMF is defined as a function that satisfies the following requirements: 1. In the whole data set, the number of extrema and the number of zero-crossings must either be equal or differ at most by one. 2. At any point, the mean value of the envelope defined by the local maxima and the envelope defined by the local minima is zero. Therefore, an IMF represents a simple oscillatory mode as a counterpart to the simple harmonic function, but it is much more general: instead of constant amplitude and frequency in a simple harmonic component, an IMF can have variable amplitude and frequency along the time axis.

Figure 16: flow chart of an EMD process.

The procedure of extracting an IMF is called sifting. The sifting process is as follows: 1. Identify all the local extrema in the test data. 2. Connect all the local maxima by a cubic spline line as the upper envelope. 3. Repeat the procedure for the local minima to produce the lower envelope.

The upper and lower envelopes should cover all the data between them. Their mean is m1. The difference between the data and m1 is the first component h1: ( ) =

Ideally, h1 should satisfy the definition of an IMF, for the construction of h1 described above should have made it symmetric and having all maxima positive and all minima negative. After the first round of sifting, the crest may become a local maximum. New extrema generated in this way actually reveal the proper modes lost in the initial examination. In the subsequent sifting process, h1 can only be treated as a proto-IMF. In the next step, it is treated as the data, then = After repeated sifting up to k times, h1 becomes an IMF, that is
( )

Then, it is designated as the first IMF component from the data:

The stoppage criteria of the sifting process The stoppage criterion determines the number of sifting steps to produce an IMF. Two different stoppage criteria have been used traditionally:

1. The first criterion is proposed by Huang et al. (1998). It similar to the Cauchy convergence test, and we define a sum of the difference, SD, as | ( ) ( ) ( )|

Then the sifting process is stop when SD is smaller than a pre-given value.

2. A second criterion is based on the number called the S-number, which is defined as the number of consecutive siftings when the numbers of zero-crossings and extrema are equal or at most differing by one. Specifically, an S-number is pre-selected. The sifting process will stop only if for S consecutive times the numbers of zero-crossings and extrema stay the same, and are equal or at most differ by one. Once a stoppage criterion is selected, the first IMF, c1, can be

obtained. Overall, c1 should contain the finest scale or the shortest period component of the signal. We can, then, separatec1 from the rest of the data by ( ) = Since the

residue, r1, still contains longer period variations in the data, it is treated as the new data and subjected to the same sifting process as described above. This procedure can be repeated to all the subsequent rj's, and the result is

The sifting process stops finally when the residue, rn, becomes a monotonic function from which no more IMF can be extracted. From the above equations, we can induce that ( )

Thus, a decomposition of the data into n-empirical modes is achieved. 4.3 EMD OF A SIGNAL

Figure 17: Original sine wave

Figure 18: Imf1

Figure 19: Imf2

Figure 20: Imf3

Figure 21: Imf4

Figure 22: Imf5

Figure 23: Imf6

figure25: figure 24: Original sine wave= imf1+ imf2+ imf3+ imf4+ imf5+ imf6.

EMD OF AN IMAGE
Empirical mode decomposition (EMD) was introduced for the analysis of time domain signals. The results and ideas in time-domain applications using EMD apply to two-dimensional (2D) signals, such as images, as well. IEMD decomposes the spatial frequency components into a set of IMFs where the highest spatial frequency component of each spatial position is in the first IMF and the second highest spatial frequency component of each spatial position is in the second IMF, and so on. An IMF is defined as a function in which the number of extrema points and the number of zero crossings are the same or differ by one. The upper and lower envelopes of the IMF are symmetric with respect to the local mean. The EMD of images relies on proper interpolation in two dimensions. The problem is to fit a surface to the 2D scattered data points representing the extrema points. The interpolated surface must go through each data point. Overshot must be avoided and the second derivative must be continuous everywhere for the signal to be smooth enough for 2D-EMD.

The residue signal should have only a few extrema points. Huang state that there should be no extrema points in the one-dimensional (1D) case and ideally we would like to have a similar criterion for the 2D case. This criterion is relaxed for the 2D case. The border constraints are even more important in two dimensions than they are in the 1D-EMD case. One of the main objections for using spline interpolation in the EMD for 2D signals such as images is that the borders cause too many problems. The set of extrema points is very sparse and since the interpolation methods only interpolate between points, the borders need special care. We proposed the trick of adding extra data points at the borders to the set of extrema points.5 These extra points are placed at the corners of the image and equally spaced around the border. Without these extra points, the areas not covered by the interpolation traverse into the image in the sifting process.

Sifting for the Two-Dimensional IMF

To find the first IMF, start with the image itself as input signal (m,n)=x(m,m) . The first index is the IMF number, l = 1, . . ., L, and the second index is the iteration number, k = 1, . . .,K, in the sifting process , m and n represent the two spatial dimensions. To find the next IMF, use the residue corresponding to the previously found IMF as input signal (m, n) = r1(m,m). The sifting process to find the IMFs of a signal x(m, n),

comprises the following steps: (1) Find the positions and amplitudes of all local maxima, and find the positions and amplitudes of all local minima in the input signal. (2) Create the upper envelope by spline interpolation of the local maxima and the lower envelope by spline interpolation of the local minima. Denote the envelopes respectively. Type of spline depends on application. (3) For each position (m, n), calculate the mean of the upper envelope and the lower envelope: (m, n) =
( ) ( )

), and

The signal (m, n) is referred to as the envelope mean. (4) Subtract the envelope mean signal from the previous signal: ( ) ( )( ) ( )

This is one iteration of the sifting process. The next step is to check if the signal step 4 is an IMF or not. The process stops when the envelope mean signal is close enough to zero . | ( )|< ( )

)from

The value of in the stop criterion affects the EMD in such a way that if it is not small enough, then there will not be a sufficient number of IMFs to separate all intrinsic modes in the signal. On the other hand, if the number were too small, the iterations will take long time. Forcing the envelope mean to zero will give us the wanted symmetry of the envelope and the correct relation between the number of zero-crossings and the number of extremes that define the IMF. This way we will find the IMF without actually having to check for symmetric envelopes.

(5) Check if the mean signal is close enough to zero, based on the stop criterion. If not, repeat the process from step 1 with the resulting signal from step 4 as the input signal a sufficient number of times. When the stop criterion is met, the IMF ( After the IMF is found, define the residue as ( ) ( ( ) )- ( ( ) . The EMD is completed when the ) ) ( ( ) is defined as the last result of (4). )

(6) The next IMF is found by starting over from step 1, now with the residue as the input signal.
( )

Steps (1) to (6) can be repeated for all the subsequent

residue, ideally, does not contain any extrema points. The signal can be expressed as the sum of IMFs and the last residue: ( ) ( ) ( )