Navigation

Related Articles

Back to Latest Articles

Histogram Equalization And Magic Behind It

C# tutorial on how to achieve histogram equalization on an image.


Andraz Krzisnik
Histogram Equalization And Magic Behind It

This tutorial is meant to demonstrate how histogram equalization works. This is a continuation on contrast stretch articles I’ve made in the past. An article on histogram equalization already exists on here, but it is made in “unsafe” code and that just kinda bothers me a little bit.

This article is accompanied by code, that manipulates bytes of an image with a “lockbits” method. This method basically stores image into a one dimensional byte array.

How does histogram equalization even work?

Histograms are charts to visualize data in columns.We can create a histogram from an image, where vertical axis shows us number of pixels and horizontal axis shows color intensities that these pixels have.

Images that need to be processed here will have a washed out look. This is because the range of intensities in this image doesn’t span the whole range of possible intensities. Histogram equalization basically stretches this range to cover complete spectrum of possible intensities.

Sounds familiar doesn’t it? We talked about the same effect in article for normalization, but the key difference here is that we use a different equation to achieve this effect. As it turns out normalization is not as effective in expanding the range as histogram equalization is.

Math behind this process

pn = number of pixels with intensity n / total number of pixels

pn represents the probability of each intensity in an image.

And finally the equation that does all the heavy lifting,

histogram equalization contrast stretch

where k represents intensity value and L all possible intensities.

For example, if we have a pixel with intensity value of 100, we need to sum all probabilities pn up to the probability for pixels with intensity value of 100. Then we need to multiply this summation by 255 and round it down to the closest integer value. Since we are working with images of 32 bit pixel format, the max intensity value for each channel is 255, hence the multiplication with this number.



Let’s take a look at the function written in C#

public static Bitmap HistEq(this Bitmap img)
        {
            int w = img.Width;
            int h = img.Height;
            BitmapData sd = img.LockBits(new Rectangle(0, 0, w, h),
                ImageLockMode.ReadOnly, PixelFormat.Format32bppArgb);
            int bytes = sd.Stride * sd.Height;
            byte[] buffer = new byte[bytes];
            byte[] result = new byte[bytes];
            Marshal.Copy(sd.Scan0, buffer, 0, bytes);
            img.UnlockBits(sd);
            int current = 0;
            double[] pn = new double[256];
            for (int p = 0; p < bytes; p += 4)
            {
                pn[buffer[p]]++;
            }
            for (int prob = 0; prob < pn.Length; prob++)
            {
                pn[prob] /= (w * h);
            }
            for (int y = 0; y < h; y++)
            {
                for (int x = 0; x < w; x++)
                {
                    current = y * sd.Stride + x * 4;
                    double sum = 0;
                    for (int i = 0; i < buffer[current]; i++)
                    {
                        sum += pn[i];
                    }
                    for (int c = 0; c < 3; c++)
                    {
                        result[current + c] = (byte)Math.Floor(255 * sum);
                    }
                    result[current + 3] = 255;
                }
            }
            Bitmap res = new Bitmap(w, h);
            BitmapData rd = res.LockBits(new Rectangle(0, 0, w, h),
                ImageLockMode.WriteOnly, PixelFormat.Format32bppArgb);
            Marshal.Copy(result, 0, rd.Scan0, bytes);
            res.UnlockBits(rd);
            return res;
        }

Demonstration

contrast stretch

Entire Project is Available for Download

Related Articles

Region Segmentation With K Means Clustering

How To Make K Means Clustering Algorithm With C#

K means clustering is a optimization method of partitioning an image by measuring Euclidean distances between pixels and cluster means.

Posted on by Andraz Krzisnik
Image Noise

How To Make Uniform Noise On Images – C# Guide

Uniform noise is one of the noise models we can use to simulate real life data corruption. This guide shows how to make in on images.

Posted on by Andraz Krzisnik