Will be the same color domain noise filtering, the implementation effect of figure 1 to figure 2,
I hope god can provide some train of thought, had better use OpenCV or can be implemented in Matlab,

CodePudding user response:

With related algorithm implementation in opencv

CodePudding user response:

Baidu search relevant keywords,

CodePudding user response:

baidu

CodePudding user response:

CodePudding user response:

reference 1st floor OExpress response:

opencv related algorithm implementation of

Please give you the details of which algorithm?

CodePudding user response:

Previous page directory page

Smooth


Smooth
Various methods of image smooth

Void cvSmooth (const CvArr * SRC, CvArr * DST,
Int smoothtype=CV_GAUSSIAN,
Int param1=3, int param2=0, double param3=0, double param4=0);
The SRC
The input image.
DST
The output image.
Smoothtype
Smoothing methods:
CV_BLUR_NO_SCALE (simple without scale transform fuzzy) - param1 x param2 field summation of each pixel, if the neighborhood size is change, can use function beforehand cvIntegral calculating integral image,
For each pixel param1 CV_BLUR (simple blur) - x param2 neighborhood sum and scale transformation 1/(param1? Param2).
CV_GAUSSIAN (gaussian blur) - nuclear size on the image for param1 x param2 gaussian convolution
CV_MEDIAN (median blur) - nuclear size of param1 x param1 on image median filtering (i.e. for neighborhood is square).
CV_BILATERAL (bidirectional filter) - application of two-way 3 x3 filter, color sigma=param1, sigma=param2 space. On the bidirectional filter, may refer to http://www.dai.ed.ac.uk/CVonline/LOCAL_COPIES/MANDUCHI1/Bilateral_Filtering.html
Param1
The first parameter to the smooth operation.
Param2
Smooth operation of the second parameter. For simple/non gaussian blur scale transform, if the value is zero param2, has said it is setting of param1,
Param3
Corresponding to the Gaussian parameters of Gaussian sigma (standard deviation). If zero, the standard deviation by the nuclear dimension calculation of the following:
Sigma=(n/2-1) * 0.3 + 0.8, where n=param1 corresponding to the level of nuclear,
N=param2 corresponding vertical nuclear.
For small convolution kernels (3 x 3 to 7 * 7) using the above formula shown in standard sigma will be fast, if param3 is not zero, and param1 and param2 is zero, the nuclear size have sigma calculation (to ensure the operation of precise enough).

CvSmooth function can use any kind of way to smooth the image above, each method has its own characteristics and limitations,

Without scaling the image smooth only supports single channel image, and support the transformation of 8 bit to 16 bit (similar to cvSobel and cvaplace) and 32-bit floating point Numbers to 32-bit floating-point format transformation,

Simple blur and gaussian blur support 1 - or 3 - channels, 8 - bit and 32 bit floating point image, the two methods can process images (in place),

Median and bidirectional filter work in 1 - or 3 - channel, 8 - bit image, but not in the in - place way processing images.


Collection time: on July 18, 2007, 21 points 12 points
Edit time: on July 18, 2007, 21 points 12 points

CodePudding user response:

Salt and pepper noise, it is recommended that the filter with middle finger, if you want to keep good edge bileteral filter can be used

CodePudding user response:

Noise belongs to the high frequency signal, you use the appropriate filter filter,

But filtering cannot produce you have shown light spot, to produce light effects, you should take photos in PCB, increase on the back of the light source, from PCB hole (transmission)

CodePudding user response:

Median filtering and the weighted average

CodePudding user response:

I think you can try to do a few times more inflation, the picture at the operation of the corrosion
CvErode corrosion, cvDilate swell

CodePudding user response:

From figure 1 to figure 2 is basically impossible, the image has been degraded, back is not so good, figure 2 unless you have a degradation model, combined with can also filter out noise,

CodePudding user response:

BM3D should have the