how can i blur the background?

Hello, I want to blur the background not the man. please help as soon as possible.

[Walter Roberson]

Note that in the general case, it is impossible for a program to automatically select which part of the image is "background" and which part is "foreground".

Consider the images produced by SOHO, the solar observatory. The images are primarily intended for observing the Sun, in which case you ignore everything outside of the solar disk. But if you take the same images and mask out the solar disk then you can observe objects moving close to the solar disk -- comets!

So in one case you ignore everything outside of the solar disk, and in the other case you ignore the solar disk, and so using the same image you can get two different purposes. The foreground for one purpose is the background for the other purpose.

Therefore it is impossible to automatically select foreground versus background: the "noise" for one purpose might easily be the "signal" for a different purpose.

https://phys.org/news/2024-03-esa-nasa-solar-observatory-5000th.html

Answer 1

[DGM]

There are several answers already. Here's the latest one. See the links therein.

https://www.mathworks.com/matlabcentral/answers/2113476-how-can-i-blur-the-background-of-an-image

Answer 2

[Image Analyst]

You can try the foreground detector in the Computer Vision Toolbox.

https://www.mathworks.com/help/vision/ref/vision.foregrounddetector-system-object.html

help foregrounddetector

--- foregrounddetector not found. Showing help for vision.ForegroundDetector instead. ---

 ForegroundDetector Detect foreground using Gaussian Mixture Models
  
    detector = vision.ForegroundDetector returns a foreground detector System 
    object, detector, that computes foreground mask using Gaussian Mixture
    Models (GMM) given a series of either grayscale or color video frames.
  
    detector = vision.ForegroundDetector('PropertyName', PropertyValue, ...) 
    returns a foreground detector System object, H, with each specified 
    property set to the specified value.
  
    Step method syntax:
  
    foregroundMask = step(detector, I) computes the foreground mask for input 
    image I, and returns a logical mask where true represents foreground
    pixels. Image I can be grayscale or color. This form of the step 
    function call is allowed when AdaptLearningRate is true (default).
 
    foregroundMask = step(detector, I, learningRate) computes the foreground 
    mask for input image I using the LearningRate provided by the user. This
    form of the step function call is allowed when AdaptLearningRate is 
    false.
  
    System objects may be called directly like a function instead of using
    the step method. For example, y = step(obj, x) and y = obj(x) are
    equivalent.
 
    ForegroundDetector methods:
  
    step     - See above description for use of this method
    reset    - Resets the GMM model to its initial state
    release  - Allow property value and input characteristics changes
    clone    - Create foreground detection object with same property values
    isLocked - Locked status (logical)
  
    ForegroundDetector properties:
  
    AdaptLearningRate          - Enables the adapting of LearningRate as 
                                 1/(current frame number) during the
                                 training period specified by 
                                 NumTrainingFrames
    NumTrainingFrames          - Number of initial video frames used for 
                                 training the background model
    LearningRate               - Learning rate used for parameter updates
    MinimumBackgroundRatio     - Threshold to determine the gaussian modes
                                 in the mixture model that constitute the
                                 background process
    NumGaussians               - Number of distributions that make up the
                                 foreground-background mixture model
    InitialVariance            - Initial variance to initialize all
                                 distributions that compose the
                                 foreground-background mixture model
  
    Example: Detect moving cars in video
    ------------------------------------
    reader = VideoReader('visiontraffic.avi');
    detector = vision.ForegroundDetector();
 
    blobAnalyzer = vision.BlobAnalysis(...
        'CentroidOutputPort', false, 'AreaOutputPort', false, ...
        'BoundingBoxOutputPort', true, 'MinimumBlobArea', 250);
  
    player = vision.DeployableVideoPlayer();
 
    while hasFrame(reader)
      frame  = readFrame(reader);
      fgMask = step(detector, frame);
      bbox   = step(blobAnalyzer, fgMask);
 
      % draw bounding boxes around cars
      out = insertShape(frame, 'Rectangle', bbox, 'ShapeColor', 'Yellow');
      step(player, out); % view results in the video player
    end
 
    release(player);
 
  See also: VideoReader, vision.BlobAnalysis, regionprops, vision.KalmanFilter, 
    imopen, imclose

    Documentation for vision.ForegroundDetector
       doc vision.ForegroundDetector

I have no idea how will its algorithm will work for your particular image(s).

Otherwise you can manually trace the foreground with drawfreehand to make a mask. Then blur the entire image and assign the blurred pixels inside the mask to the original unblurred image. Like for once you have the mask:

h = ones(15);
filteredRGB = imfilter(originalRGB, h);
[r,g,b] = imsplit(rgbImage);
[rf, gf, bf] = imsplit(filteredRGB);
r(mask) = rf(mask);
g(mask) = gf(mask);
b(mask) = bf(mask);
outputImage = cat(3, r, g, b);

See attached drawing demos for creating and masking images.