Solucion Planteada

To solve the problem, an algorithm is presented that consists of a series of operations applied to the images, both to clean them and to enhance their characteristics, perform a subtraction operation to detect the movement of the person and once the area of interest will proceed to define it more clearly and then fill in the resulting area with the image of the person. In this way the surrounding border is eliminated and we obtain the desired results.

Some of the reasons why the operations presented and the structure of the program developed are indicated below.

• GaussianBlur (5x5) 
• The Gaussian filter was chosen because it presents better results to soften edges within the images, with the smoothed edges we obtain two results, one that helps us eliminate the noises that the images may have and another that reduces the possibility of detecting them as objects of interest other things that are not people.
• Color space Gray
• This color space is chosen because it is going to work with morphological operations and with threshold binarization. This color space is chosen because it is going to work with morphological operations and with threshold binarization.
• Top-Hat, Black Hat morphological operation
• This operation allows us to obtain new characteristics in an image such as edges that are not easily detected. The application of these operations and then applying operations such as addition and subtraction to them allow us to obtain new results, in the case of practice it allowed us to better define the edges of the person.
• Images subtraction
• With the images treated with morphological operations, we make a subtraction where we only have the place where the person is moving, that way we have an area of interest where we will work.
• Morfological operation Open
• With the subtraction of images, such as videos and for reasons of light or wind, small objects that move in the scene can be presented and this is seen as small noises of salt in the image, for this we use the Open operation that helps us eliminate those points.
• Binarization Umbral
• Once we are left with the area of interest, we apply a threshold binarization to stay with the pixels of interest and convert them to white so that it is easier to work with that area.
• Morphological operation Dilate y Close
• With these operations what is sought is to improve the area of interest, with the dilate operation we increase the size of the edges of the image and with the close operation we fill in the interior spaces so we are left with a white spot that represents the person in motion .