平滑锯齿状路径

前几天,我参与了“图像/图形进入形状”线程,并通过迭代地将图像添加到.这是非常缓慢的。RectangleArea

此示例改为生成 一个,并从 GP 创建 。快得多。GeneralPathArea

Sample images for processing

左上角的图像是“源图像”。右边的两个是处理大纲的各个阶段。它们两者在圆圈周围和沿着三角形的倾斜边都有锯齿状的边缘。

我想获得一个去除或减少锯齿的形状。

在 ASCII 艺术中。

案例1:

  1234
1 **
2 **
3 ***
4 ***
5 ****
6 ****

拐角处:

  • (2,3) 内角
  • (3,3)
  • (3,5) 内角
  • (4,5)

案例2:

  1234
1 ****
2 ****
3 **
4 **
5 ****
6 ****

拐角处:

  • (4,2)
  • (2,2) 内角
  • (2,5) 内角
  • (4,5)

假设我们的路径显示了形状和列出的点,我想删除第一组的“内角”点,同时保留“一对”内角(从图像中咬出一口)作为第二组。


  • 任何人都可以建议一些聪明的内置方法来完成这项工作的繁重工作吗?
  • 如果做不到这一点,那么识别内角的位置和性质(对/单)的好方法是什么?(我猜我可以得到一个并建立一个新的下降单数的内角 - 只要我能弄清楚如何识别它们!PathIteratorGeneralPath

下面是要使用的代码:

import java.awt.*;
import java.awt.event.*;
import java.awt.image.*;
import java.awt.geom.*;
import javax.swing.*;
import javax.swing.border.*;
import javax.swing.event.*;

/* Gain the outline of an image for further processing. */
class ImageOutline {

    private BufferedImage image;

    private TwoToneImageFilter twoToneFilter;
    private BufferedImage imageTwoTone;
    private JLabel labelTwoTone;

    private BufferedImage imageOutline;
    private Area areaOutline = null;
    private JLabel labelOutline;

    private JLabel targetColor;
    private JSlider tolerance;

    private JProgressBar progress;
    private SwingWorker sw;

    public ImageOutline(BufferedImage image) {
        this.image = image;
        imageTwoTone = new BufferedImage(
            image.getWidth(),
            image.getHeight(),
            BufferedImage.TYPE_INT_RGB);
    }

    public void drawOutline() {
        if (areaOutline!=null) {
            Graphics2D g = imageOutline.createGraphics();
            g.setColor(Color.WHITE);
            g.fillRect(0,0,imageOutline.getWidth(),imageOutline.getHeight());

            g.setColor(Color.RED);
            g.setClip(areaOutline);
            g.fillRect(0,0,imageOutline.getWidth(),imageOutline.getHeight());
            g.setColor(Color.BLACK);
            g.setClip(null);
            g.draw(areaOutline);

            g.dispose();
        }
    }

    public Area getOutline(Color target, BufferedImage bi) {
        // construct the GeneralPath
        GeneralPath gp = new GeneralPath();

        boolean cont = false;
        int targetRGB = target.getRGB();
        for (int xx=0; xx<bi.getWidth(); xx++) {
            for (int yy=0; yy<bi.getHeight(); yy++) {
                if (bi.getRGB(xx,yy)==targetRGB) {
                    if (cont) {
                        gp.lineTo(xx,yy);
                        gp.lineTo(xx,yy+1);
                        gp.lineTo(xx+1,yy+1);
                        gp.lineTo(xx+1,yy);
                        gp.lineTo(xx,yy);
                    } else {
                        gp.moveTo(xx,yy);
                    }
                    cont = true;
                } else {
                    cont = false;
                }
            }
            cont = false;
        }
        gp.closePath();

        // construct the Area from the GP & return it
        return new Area(gp);
    }

    public JPanel getGui() {
        JPanel images = new JPanel(new GridLayout(2,2,2,2));
        JPanel  gui = new JPanel(new BorderLayout(3,3));

        JPanel originalImage =  new JPanel(new BorderLayout(2,2));
        final JLabel originalLabel = new JLabel(new ImageIcon(image));
        targetColor = new JLabel("Target Color");
        targetColor.setForeground(Color.RED);
        targetColor.setBackground(Color.WHITE);
        targetColor.setBorder(new LineBorder(Color.BLACK));
        targetColor.setOpaque(true);

        JPanel controls = new JPanel(new BorderLayout());
        controls.add(targetColor, BorderLayout.WEST);
        originalLabel.addMouseListener( new MouseAdapter() {
            @Override
            public void mouseEntered(MouseEvent me) {
                originalLabel.setCursor(
                    Cursor.getPredefinedCursor(Cursor.CROSSHAIR_CURSOR));
            }

            @Override
            public void mouseExited(MouseEvent me) {
                originalLabel.setCursor(Cursor.getDefaultCursor());
            }

            @Override
            public void mouseClicked(MouseEvent me) {
                int x = me.getX();
                int y = me.getY();

                Color c = new Color( image.getRGB(x,y) );
                targetColor.setBackground( c );

                updateImages();
            }
        });
        originalImage.add(originalLabel);

        tolerance = new JSlider(
            JSlider.HORIZONTAL,
            0,
            255,
            104
            );
        tolerance.addChangeListener( new ChangeListener() {
            public void stateChanged(ChangeEvent ce) {
                updateImages();
            }
        });
        controls.add(tolerance, BorderLayout.CENTER);
        gui.add(controls,BorderLayout.NORTH);

        images.add(originalImage);

        labelTwoTone = new JLabel(new ImageIcon(imageTwoTone));

        images.add(labelTwoTone);

        images.add(new JLabel("Smoothed Outline"));

        imageOutline = new BufferedImage(
            image.getWidth(),
            image.getHeight(),
            BufferedImage.TYPE_INT_RGB
            );

        labelOutline = new JLabel(new ImageIcon(imageOutline));
        images.add(labelOutline);

        updateImages();

        progress = new JProgressBar();

        gui.add(images, BorderLayout.CENTER);
        gui.add(progress, BorderLayout.SOUTH);

        return gui;
    }

    private void updateImages() {
        if (sw!=null) {
            sw.cancel(true);
        }
        sw = new SwingWorker() {
            @Override
            public String doInBackground() {
                progress.setIndeterminate(true);
                adjustTwoToneImage();
                labelTwoTone.repaint();
                areaOutline = getOutline(Color.BLACK, imageTwoTone);

                drawOutline();

                return "";
            }

            @Override
            protected void done() {
                labelOutline.repaint();
                progress.setIndeterminate(false);
            }
        };
        sw.execute();
    }

    public void adjustTwoToneImage() {
        twoToneFilter = new TwoToneImageFilter(
            targetColor.getBackground(),
            tolerance.getValue());

        Graphics2D g = imageTwoTone.createGraphics();
        g.drawImage(image, twoToneFilter, 0, 0);

        g.dispose();
    }

    public static void main(String[] args) throws Exception {
        int size = 150;
        final BufferedImage outline =
            new BufferedImage(size,size,BufferedImage.TYPE_INT_RGB);
        Graphics2D g = outline.createGraphics();
        g.setColor(Color.WHITE);
        g.fillRect(0,0,size,size);
        g.setRenderingHint(
            RenderingHints.KEY_DITHERING,
            RenderingHints.VALUE_DITHER_ENABLE);
        g.setRenderingHint(
            RenderingHints.KEY_ANTIALIASING,
            RenderingHints.VALUE_ANTIALIAS_ON);

        Polygon p = new Polygon();
        p.addPoint(size/2, size/10);
        p.addPoint(size-10, size-10);
        p.addPoint(10, size-10);
        Area a = new Area(p);

        Rectangle r = new Rectangle(size/4, 8*size/10, size/2, 2*size/10);
        a.subtract(new Area(r));

        int radius = size/10;
        Ellipse2D.Double c = new Ellipse2D.Double(
            (size/2)-radius,
            (size/2)-radius,
            2*radius,
            2*radius
            );
        a.subtract(new Area(c));

        g.setColor(Color.BLACK);
        g.fill(a);

        ImageOutline io = new ImageOutline(outline);

        JFrame f = new JFrame("Image Outline");
        f.setDefaultCloseOperation(JFrame.EXIT_ON_CLOSE);
        f.add(io.getGui());
        f.pack();
        f.setResizable(false);
        f.setLocationByPlatform(true);
        f.setVisible(true);
    }
}

class TwoToneImageFilter implements BufferedImageOp {

    Color target;
    int tolerance;

    TwoToneImageFilter(Color target, int tolerance) {
        this.target = target;
        this.tolerance = tolerance;
    }

    private boolean isIncluded(Color pixel) {
        int rT = target.getRed();
        int gT = target.getGreen();
        int bT = target.getBlue();
        int rP = pixel.getRed();
        int gP = pixel.getGreen();
        int bP = pixel.getBlue();
        return(
            (rP-tolerance<=rT) && (rT<=rP+tolerance) &&
            (gP-tolerance<=gT) && (gT<=gP+tolerance) &&
            (bP-tolerance<=bT) && (bT<=bP+tolerance) );
    }

    public BufferedImage createCompatibleDestImage(
        BufferedImage src,
        ColorModel destCM) {
        BufferedImage bi = new BufferedImage(
            src.getWidth(),
            src.getHeight(),
            BufferedImage.TYPE_INT_RGB);
        return bi;
    }

    public BufferedImage filter(
        BufferedImage src,
        BufferedImage dest) {

        if (dest==null) {
            dest = createCompatibleDestImage(src, null);
        }

        for (int x=0; x<src.getWidth(); x++) {
            for (int y=0; y<src.getHeight(); y++) {
                Color pixel = new Color(src.getRGB(x,y));
                Color write = Color.BLACK;
                if (isIncluded(pixel)) {
                    write = Color.WHITE;
                }
                dest.setRGB(x,y,write.getRGB());
            }
        }

        return dest;
    }

    public Rectangle2D getBounds2D(BufferedImage src) {
        return new Rectangle2D.Double(0, 0, src.getWidth(), src.getHeight());
    }

    public Point2D getPoint2D(
        Point2D srcPt,
        Point2D dstPt) {
        // no co-ord translation
        return srcPt;
    }

    public RenderingHints getRenderingHints() {
        return null;
    }
}

答案 1

这是一个大问题。你可能会发现Johannes Kopf和Dani Lischinski的Depixelizing Pixel Art1很有用:它是可读的,最近的,包括以前工作的摘要,并详细解释了他们的方法。

另请参阅涵盖类似背景视频的幻灯片(!)。

  1. 以下是“最近邻居”与“他们的技术”文档中的一些屏幕截图。nearest neighbortheir result

答案 2

此问题的最一般版本是大多数计算机视觉管道中的初始阶段之一。它被称为图像分离。它将图像拆分为被视为视觉上相同的像素区域。这些区域由“轮廓”分隔(例如,请参阅本文),这相当于通过图像沿像素边界运行的路径。

有一个简单的递归算法,用于将轮廓表示为多段线定义,使得其中没有点偏离超过您可以选择的某个固定量(例如)。通常为 1/2 到 2 像素。max_dev

function getPolyline(points [p0, p1, p2... pn] in a contour, max_dev) {
  if n <= 1 (there are only one or two pixels), return the whole contour
  Let pi, 0 <= i <= n, be the point farthest from the line segment p0<->pn
  if distance(pi, p0<->pn) < max_dev 
    return [ p0 -> pn ]
  else
    return concat(getPolyline [ p0, ..., pi ],  getPolyline [ pi, ..., pn] )

这背后的想法是,你似乎有卡通般的图像已经被分割了。因此,如果您编写一个简单的搜索,将边缘像素组装成链,则可以使用上面的算法将它们转换为平滑的线段链。它们甚至可以用抗锯齿绘制。


推荐