Application of convolution in computer vision

 Application of convolution in computer vision(like   image processing)

-, Vinish Soni, Sarvesh Wadi , Kaustubh Wakodkar  , Yash Shamnani , Sameeran zingre

In general when we are studying about image processing we must know some basic definition like kernel,convolution etc.which is used in image processing. Starting with kernel.

Kernel :  A convolution lets you do many things, like calculate derivatives, detect edges, apply blurs, etc. A very wide variety of things. And all of this is done with a "convolution kernel".

The convolution kernel is a small matrix. This matrix has numbers in each cell and has an anchor point

This kernel slides over an image and does its thing. The "anchor" point is used to determine the position of the kernel with respect to the image. The anchor point is generally the center point.

Different image processing has different kernels for ex:

Convolution in image processing : In image processing convolution of matrix involves laying a matrix over another and  then calculate weighted sum of all pixel values. So considering below example we have a image convoluted with kernel  getting an output image.all values in image are the vibrance values of image.when we are convoluting this image at center (3,3).what we do is placing a kernel at(3,3) then we map the cells of image with the corresponding weights in the kernel so 40 maps to 0 42 to 1, 46 to 0 in similar fashion we map all the cells of the image with their weights on the kernel and then we calculate sum in this case it is 42 so what this kernel does is move the image down by one pixel this is a shift kernel in action.

Problematic corners and edges :   The kernel is two dimensional. So you have problems when the kernel is near the edges or corners.  Do something about the edges Usually people choose to do something about it. They create extra pixels near the edges. There are a few ways to create extra pixels:

• ·       Set a constant value for these pixels
• ·       Duplicate edge pixels
• ·       Reflect edges (like a mirror effect)
• This usually fixes the problem that might arise.

                            IMAGE PROCESSING       

Image processing in spatial domain is a visually rich area of study dealing with pixel-manipulation techniques. Different operations are performed over the images, which are treated simply as two-dimensional arrays.

So images are just data and data can be manipulated easily and so can images. In simple terms image processing is simply a method to extract some information from an image or insert some information in it. A common reason to process images is to make them more appealing.

Examples of image processing:

MONOCHROMATIC  GAUSSIAN  BLUR :

Every kernel is made for specific operation like gaussian blur kernel  will make whole image blur except center you can see in below ex.

    

                      →                 →       

We are seeing here monochromatic gaussian blur we have taken sample of image having 5 cross 5 grid of white pixels and 3 cross 3 grid of blacks these values are vibrance values zero for black and hundred for white when we pass this image from gaussian blur kernel we get output image seen on right and the vibrance values of image are shown in grid at bottom as you can see the vibrance value of no pixel is hundred because the black from the center has creeped out and affected all white surrounding pixels because the black grid which was earlier focussed has now defocussed by blurring we can see that only the central pixel is unaffected it remains zero as before so the convolution does not affect it as it affect the surrounding pixels.which is the work of gaussian blur kernel.

Similarly these procedure is followed in all processes depending on kernel functionality like-

Simple box blur :

This convolution kernel has an averaging effect. So you end up with a slight blur. The image convolution kernel is

Note that the sum of all elements of this matrix is 1.0. This is important. If the sum is not exactly one, the resultant image will be brighter or darker.

Here's a blur that I got on an image:

Edge detection :

The above kernels are in a way edge detectors. Only thing is that they have separate components for horizontal and vertical lines. A way to "combine" the results is to merge the convolution kernels. The new image convolution kernel looks like this:

Below result I got with edge detection:

Thus conclusion is convolution is very helpful in image processing And image processing has many real world application like camera app on mobile phones nowadays we use this in our Instagram filters,snapchats,face recognition etc.