Box (Mean) Filter

But what the heck is Mean filter? will it tell me to shut up? Well no. You might be familiar with the mathematical mean, which looks like so

$$\bar{x} = \frac{1}{n} \sum_{i=1}^{n} x_i$$

or a less obtuse way to imply the same

SUM(array)/LENGTH(array)

Mean filter works by assigning the current pixel to the mean of all its surrounding pixels, which effectively results in a blurred effect.

Implementation

The Box (Mean) filter is implemented at src/smoothing/box.c, Let's take a deeper look at it.

void cv_apply_box_filter(Image * img, int kernSize) {
  /* ... */

  for (int i = 0; i < height; i++) {
	for (int j = 0; j < width; j++) {
  	unsigned char R = compute_mean_for_channel(img, kernSize, j, i, 0);
  	unsigned char G = compute_mean_for_channel(img, kernSize, j, i, 1);
  	unsigned char B = compute_mean_for_channel(img, kernSize, j, i, 2);

  	tempBytes[(i * width + j) * ch + 0] = R;
  	tempBytes[(i * width + j) * ch + 1] = G;
  	tempBytes[(i * width + j) * ch + 2] = B;
	}
  }

  /* ... */
}

As expected we loop through every pixel, more specifically every channel within the image, and for each we compute their mean, we can do via a simple function

unsigned char compute_mean_for_channel(Image * img, int size, int x, int y, int c) {

  /* ... */

  for (int i = 0; i < size; i++) {
	for (int j = 0; j < size; j++) {
  	sum += img->bytes[((y + i) * img->width + j + x) * img->channels + c];
  	count++;
	}
  }

  /* ... */

  return sum / count;
}

My apologies if this code isn't very clear, all we doing is for the index provided, we sum up the next N values to it's right and below (where N = kernSize = size).

Result

.\bin\cv.exe mean --kernel 9 .\data\img1.jpg .\img1-mean-9.png

Original

Mean filter of Kernel size 9

Problems

And that is it. So are we done? Is the problem of blurring things solved? No, this filter has some shortcomings, notably.

it can't distinguish sharper features within the image as relevant, it instead applies a uniform blur over the entire image

Which is why we need the Gaussian Blur let's see it further next.