This article refers to the address: http://
Fundamental
The image is affected by acquisition and storage, processing, and various disturbances, and noise appears on the screen during display. In order to reduce the negative effects of noise and restore the real picture as much as possible, the noise reduction filter is needed to process the image data.
There are many types of noise in an image. Such as additive noise, multiplicative noise, salt and pepper noise, Gaussian noise. Common methods of noise reduction are performed in the spatial domain, and the image data is transferred to the frequency domain after being transformed by Fourier transform. The filtering in the frequency domain needs to involve complex domain conversion operations, and the hardware implementation will consume more resources and time. Methods performed in the spatial domain include mean or weighted mean filtering, median or weighted median filtering, minimum mean squared filtering, and multiple iterations of mean or median.
Practice has proved that although these methods have certain noise reduction effects, they all have their limitations. For example, the weighted mean is very obvious in terms of detail loss; while the median is only effective for pulse interference, but not for Gaussian noise.
In fact, image noise is always intertwined with valid data. If not handled properly, the boundary contours, lines, etc. will be blurred, which will reduce the image quality. In addition, since various types of noise often do not exist alone but coexist at the same time, simply using a filter often fails to achieve satisfactory results.
The algorithm described in this paper is aimed at this problem, and proposes a new digital image filter method that can filter out multiple noises at the same time. The filtering effect is better than that of a simple filter, and the calculation amount is not too small. Complex, easier to implement on hardware.
Algorithm Description
The filtering algorithm used in this paper includes the following main steps.
â— Select multiple neighborhood pixels fi in a neighborhood window, (i=1, 2,...,n);
â— Noise detection: The number of differences between the accumulated center pixel and other pixels in the neighborhood is greater than the number of T1, and T1 is a threshold related to the degree of detail of the image;
â— Calculate the multi-window corresponding to the selected pixel (including three types of window types: cross-shaped window, X-shaped window and rectangular window). The intermediate value after sorting and comparing the pixel data is represented by “g2â€.
â— Calculate the final output value of the filter according to equation (1) to adjust the pixel data of the center pixel.
(1)
In the above formula, g(x, y) represents the output pixel value of the filter; a and b are the linear and nonlinear filtering scale coefficients selected for different noise types, respectively, and their values ​​are related to the magnitude of N.
Experimental result
In order to compare the noise reduction effect of the proposed method and the original method, we take the Lena diagram as an example, adding a Gaussian noise with a mean of 0.003, a difference of 0.004, and an impulse noise of 0.01, and then using weighted mean filtering. The weighted median filtering and the method proposed in this paper filter the noise figure. The effect is shown in Figure 1.
Figure 1 (a) Lena original map
Figure 1(b) shows the Lena diagram of Gaussian and salt and pepper noise.
Figure 1(c) After weighted mean filtering
Figure 1 (d) after weighted median filtering
Figure 1 (e) after filtering the method of this article
At the same time, the method of calculating the signal-to-noise ratio can be used to make a quantitative comparison of various methods. Equation (2) is the calculation formula of the signal to noise ratio.
(2)
Where M and N are the length and width of the image, g is the filtered data, s is the original image data, and f is the image data to which the noise is added. If the calculated signal-to-noise ratio R is negative, the noise is suppressed, and the smaller the value of R, the better the filtering effect. Table 1 shows the signal to noise ratio results for various methods.
It can be seen that the method described in the present invention can combine the advantages of mean filtering for effective Gaussian noise and median filtering for impulse noise, while avoiding the situation of blurring details while filtering. Therefore, experiments have proved that the method described in this paper is effective.
Hardware implementation
Figure 2 is a block diagram of the hardware system. A large number of tests have shown that the human eye is more sensitive to luminance signals than to chrominance. It is specially designed for noise reduction filtering of the brightness signal, which is in line with the human eye's viewing mode. Therefore, the RGB signal stream is first subjected to YUV color space conversion for bright color separation. The RAM block completes the bright and color data storage of the n-1 line, which can realize the input and output of the neighborhood window data stream, output to the subsequent noise judgment and filter processing module, and finally restore to the RGB output to the display terminal.
Figure 2 hardware system block diagram
Key device selection should consider the system operating frequency, total module size, calling IP library, chip cost, scalability and scalability, vendor support and other factors. Since this platform is only used for the module verification phase in the chip development phase, it is not necessary to choose the best performing and expensive series. In combination with the actual situation of this design, Xilinx's Spartan 3E series XC3S500E (PQ208 package) is selected. The chip contains 20 RAM blocks, which meets the requirement of 14 RAMs for storing two rows of real-time pixel data and corresponding control signals. 10476 logical units, 4656 slices, and the system is synthesized by Synplify with a scale of 3158 slices, accounting for about 70% of internal resources, leaving a certain margin. Figure 3 is a flow chart of the main filter module.
Figure 3 Flow chart of the main module of the filter
Considerations in circuit design
According to the Video Electronics Standards Association (VESA), at least a frequency of about 135 MHz is required for the pixel clocks of video SD and HD signals. In order to meet this frequency requirement, it is necessary to apply the flow pipeline design method to comprehensively consider the factors of both area and speed.
There are two difficulties in design: one is how to build a real-time pixel matrix data stream, and simultaneously output n×n matrix data at a time; the second is how to use the pipeline method to output the filtered pixels in real time after delaying a few beats. data.
For the first problem, you can call the RAM core provided by Xilinx, and use the Coregen tool to plan a storage space of n rows or n-1 rows inside the device. After the nth data of the nth row arrives, it can output simultaneously. n × n matrix data to the subsequent processing module. After outputting a matrix, the new data enters the storage space again, causing the data to be refreshed. Taking the 3×3 matrix as an example, the function simulation waveform of the memory module is shown in Figure 4. It can be seen that after delaying one row plus one pixel clock, the module starts to output data in nine neighborhoods from y11 to y33 by clock.
Figure 4 Functional simulation waveform of the memory module
The solution to the second problem can make full use of the pipeline method that can effectively increase the speed of the circuit. Firstly, the filtering process is divided into several steps, each step corresponding to some status flag registers or intermediate registers, thereby constructing a pipeline structure with several beats, and finally real-time after the total number of delays (ie, by image pixel clock) Frequency to output) filtered data output stream. Whether it is a 3×3 or 7×7 matrix neighborhood, or a larger neighborhood, the intermediate value from the input matrix data to the output matrix can be completed with a maximum of 8 beats. Therefore, the operating frequency of the circuit is effectively improved. The result of the system after static timing analysis is 135 MHz, which satisfies the pixel clock frequency required for high-definition signal processing.
What suddenly appear into your mind when hear Mini Gaming PC? Is there quality heat-releasing fan and box design of Mini Gaming PC Build? Does it run stable and smoothly when handle heavier jobs, like Photoshop, Pr, engineering design and drawing, 3d Max, big games, etc. ? now we can be responsible to answer [ Yes". No matter cheap mini gaming pc or Mini Gaming PC Under $500, equips with quality fan to heat releasing. Therefore, no need to worry that again.
To processor, can do from intel celeron j4125, N5105 up to i3, i5 i7 10th 11th 12th with or without video graphics. To slots, same rich as traditional computer tower, like above 4 or 6 usb ports, 2 Rj45, VGA, PD, slot, etc.
You can also find Mini Gaming PC, j4125 Mini PC ,Custom All In One PC, windows 10 Education Laptop,Quad Core Processor Laptop, 15 Inch Gaming Laptop ,windows Yoga Laptop , 8 inch Android Tablet, etc.
If any other special requirements interest, you can also let us know, will try our best support you.
Meet your unique demand in this field is our mission, so just feel free contact us whenever you have different idea.
Mini Gaming PC Build,Cheap Mini Gaming PC,Mini Gaming PC Under $500,Mini Gaming Desktop
Henan Shuyi Electronics Co., Ltd. , https://www.shuyioemminipc.com