Question

Scope:

I am trying to “Binarize” ( convert to only two colors ) the images that can be found here. The binarized output must follow this logic : Characters will be painted as a “Color X” and everything else will be painted as “Color Y”.

Ideally X is White and Y is black( but X and Y does not really matter).

Observations:

After some observations, i figured out that the characters are either “Yellow tones”, “White Tones” or “Black Tones”.
Using the code Bellow, i should be able to recognize and convert all those “tones” (Color channels thresholding) to one color, and whole rest, into another color.

Code Sample:

// My Own Web Requests Class. You can use C# 
WebRequests wr = new WebRequests();  

Native One
Bitmap bmp;
string url = "http://www.fazenda.rj.gov.br/projetoCPS/codigoImagem";

bmp = wr.GetBitmap(url);
bmp.Save(@"captcha.bmp");

for (int i = 0; i < bmp.Width; i++)
{
    for (int j = 0; j < bmp.Height; j++)
    {
        Color pixel = bmp.GetPixel(i, j);
        byte R, G, B;

        // Fetching RGB Values
        R = pixel.R;
        G = pixel.G;
        B = pixel.B;

        // Converting Black tones 
        if ((R < Color.Black.R + 110)
            && (G < Color.Black.G + 110)
            && (B < Color.Black.B + 110))
        {
            bmp.SetPixel(i, j, Color.Orange);
            continue;
        }

        // Converting Yellow tones
        if ((R > Color.Yellow.R - 110)
                && (G > Color.Yellow.G - 110)
                && (B < Color.Yellow.B + 110))
        {
            bmp.SetPixel(i, j, Color.Orange);
            continue;
        }

        // Converting White tones
        if ((R > Color.White.R - 110)
            && (G > Color.White.G - 110)
            && (B > Color.White.B - 110))
        {
            bmp.SetPixel(i, j, Color.Orange);
            continue;
        }

        bmp.SetPixel(i, j, Color.Orchid);
    }
}

bmp.Save(@"FilteredCaptcha.bmp");

Problem:

After running this code, the saved output (Bitmap) actually have more than the two colors i defined in code (which are “Orange” for characters and “Orchid” for everything else, in this sample).

Once i “Zoom In” using Windows Explorer / Gimp / ImageProcessingLab , there are “tones” of orange and “tones” of orchid.

What am i missing here ? Why aren’t the outputs binarized if the only uses “two” colors, instead of “tones” to Set the pixel Colors ?

I uploaded some images to make your inspection job easier.

Input Bitmap:

Output Bitmap:

Answer 1

Change where you save it to:

bmp.Save(@"c:\temp\FilteredCaptcha.bmp", System.Drawing.Imaging.ImageFormat.Bmp);

Your code is clearly only setting two colours so the only possibility I see is that the source image is a JPG, and this format is being preseved when you save it.