We build the system by using MATLAB as MATLAB is a high-level computing language. It has a variety of APIs for image processing that makes our task simple and effective.
Here the image is taken froim a scanner. Our aim is to set it to 600 DPI . DPI is described as the quantity of pixels per unit area. The smaller dpi means the lower-resolution scans performed. In any other case higher-resolution scans performed. We know that the size of an A4 size paper is 21cm × 29.7cm, 600 dpi implies that each inch contain 600 pixels. So after filtering, we get the size of the picture as 7016 × 4961 pixels. To reduce the calculation, the framework will reset the first picture to 1024×768 pixels. This is done during pre-processing.
Indian Rupees (INR) is represented as Figure 7, and here the white region is removed after the pre-preprocessing.
The system then asks the user to take the image of the currency and then the system starts the process to recognize the currency. When this recognition process starts , the system then starts with the process of image processing.If the information exhibited by the image is lost due to surface damage,noise level or sharpness issues then the recognition will be failed and the user needs to do the processing again.
The framework needn’t bother with additional gadget, our calculation depends on visual highlights for acknowledgment. It can acknowledge the money, and print out the outcome by text. The image of Indian Rupees we take by a scanner.
The UI is for printing out the first picture and its special features.
The framework can peruse JPEG (JPG) group however others. Our picture was acquired from a scanner. As referenced previously, the goals is set to 600 DPI. However, this will make the picture a major size. So subsequent to perusing in the picture, the framework will reset the picture to measure 1024 by 768 pixels and this work will allude to picture pre-processing.
The point of picture pre-processing is to stifle undesired twists or improve some picture includes that are significant for additional analysis or examination. In our work, picture pre-processing incorporates these parts:
At the point when we get the picture from a scanner, the size of the picture is so enormous. So as to diminish the figuring, we decline the size to 1024×768 pixels.
When utilizing a computerized camera or a scanner and perform picture moves, some clamor will show up on the picture. Noise in the image is due to random brightness in images. Evacuating the noise is a significant advance when picture handling is being performed. Anyway noise may influence segmentation and pattern matching.
When performing smoothing process on a pixel, the neighbor of the pixel is utilized to do some changing. After that another estimation of the pixel is made. The neighbor of the pixel is comprising with some different pixels and they develop a network, the size of the grid is odd number, the objective pixel is situated on the center of the framework.
Convolution is utilized to perform picture smoothing. As the initial step, we focus our channel over pixel that will be sifted. The channels coefficients are duplicated by the pixel esteems underneath and the outcomes are included. Then the focal pixel value as shown in the figure is changed to the new determined value.
Now at the end , the filter is again moved to the next pixel and again the convolution process takes place. These newly calculated values are not used in the next pixel filtering.
When the filter is focussed over a pixel with the border, some parts of it will lie outside the edge of the image. The techniques to handle these situations are as follows:
- Zero padding: all filter values that are outside the image are set to 0.
- Wrapping: all the filter values that lie outside the border of the image are set to its “reflection value”
- Then we start the convolution from second row and column.
- The remaining unfiltered rows and columns will get copied to the resulting image.
Here we use Gaussian operator to blur an image and minimize the noise. As mentioned in equation 2 , we create elements in Gaussian.
But there are some other ways to smooth the image, such as median filter. Median filtering is a nonlinear operation and this is often used in image processing to reduce noise. Also the median filter is more effective than convolution when the goal is to simultaneously reduce noise and preserve the edges.
After the processing is done using the median filter, the noise is removed and some detail is described on the image. The pattern which is the most important thing for our process is also clear.
After removing the noises our next step is to cut off the useless area. Sometimes, some black lines will appear on the edge of the original image, which affects the next operation. To avoid this kind of problem, we cut each side by 10 pixels.
When contrasted with an A4 size paper, the cash is little. Be that as it may, when we get the picture from the scanner, at that point the picture we get is an image like an A4 paper. So after the checking is done, the picture will have loads of white zone encompassing the cash. All things considered this is a futile part for acknowledgment. So as to make the framework effective, the white region part will be altogether cut.
Due to the light condition, when we get the picture from an advanced camera, we have to play out the histogram leveling.
Histogram leveling is utilized to modify the complexity and the brilliance of the picture, since some piece of the acknowledgment depends on shading handling. Distinctive light conditions influences the outcome. So there is a need to perform histogram evening out .
To do the procedure of division, we have to expel more things that are not expected by binarizing the picture. We needed to set the limit to choose which one is set to ‘0’ (dark) and which one is set to ‘1’ (white). As a matter of fact, the thing is there is we no compelling reason to set ‘1’. We set two qualities for the limit, the estimation of the pixel between those two qualities is set to ‘0’, and others is set to ‘1’. After commonly of testing, we set the edge somewhere in the range of 0.50 and 0.60 for the Indian Rupees .
