有時為強化影像之輪廓可以採用"Sobel 邊緣偵測",強化影像邊緣。
以3*3的 「遮罩」為例,透過「Spatial Convolution」對整張影像進行處理。
先針對已向進行"Vertical Sobel"強化垂直邊緣,在對影像進行"Horizontal Sobel"強化水平邊緣後將兩張已向合併即可取得完整之邊緣偵測影像。
亦可針對實際應用場景僅採用 Horizontal/Vertical 其中一種。
針對影像進行"Sobel 邊緣偵測"實驗成果如下圖所示 :
C#
private static Bitmap sobel(Bitmap bitmap)
{
int width = bitmap.Width, height = bitmap.Height;
int w = 3, h = 3;
Bitmap dstBitmap = new Bitmap(bitmap);
byte[,] pix = ImageExtract.getimageArray(bitmap);
byte[,] resPix = new byte[3, width * height];
for (int y = 1; y < (height - 1); y++)
{
for (int x = 1; x < (width - 1); x++)
{
//b,g,r
int current = x + (y * width);
for (int c = 0; c < 3; c++)
{
//mask
byte[] mask = new byte[w * h];
for (int my = 0; my < h; my++)
for (int mx = 0; mx < w; mx++)
{
int pos = current + (mx - 1) + ((my - 1) * width);
mask[mx + my * w] = pix[c, pos];
}
resPix[c, current] = sobelMask33(mask);
//resPix[c, current] = pix[c, current];
}
}
}
ImageExtract.writeImageByArray(resPix, dstBitmap);
return dstBitmap;
}
private static byte sobelMask33(byte[] gate)
{
int[] mask1 ={
-1,0,1,
-2,0,2,
-1,0,1
};
int[] mask2 ={
-1,-2,-1,
0,0,0,
1,2,1
};
int gay1 = 0, gay2 = 0;
for (int i = 0; i < gate.Length; i++)
{
gay1 += (gate[i] * mask1[i]);
gay2 += (gate[i] * mask2[i]);
}
int value = (int)Math.Pow((gay1 * gay1 + gay2 * gay2), 0.5);
return (byte)((value > 255) ? 255 : value);
}
完整程式碼 : https://github.com/Lung-Yu/ImageToolBox/blob/master/ImageProcessToolBox/Sobel.cs
文章標籤
全站熱搜
留言列表