Table of Contents

Thresholding, Masks, and Morphology

This example introduces OpenCV thresholding and morphology operations for creating and refining binary masks. In OpenCvSharp, a conventional mask is a single-channel CV_8UC1 Mat: zero excludes a pixel and any non-zero value includes it.

Create a binary image with a fixed threshold

Convert color input to grayscale before applying a simple threshold:

using var source = Cv2.ImRead("input.jpg", ImreadModes.Color);
if (source.Empty())
{
    throw new IOException("Could not read input.jpg.");
}

using var grayscale = new Mat();
using var binary = new Mat();

Cv2.CvtColor(source, grayscale, ColorConversionCodes.BGR2GRAY);
Cv2.Threshold(
    grayscale,
    binary,
    thresh: 128,
    maxval: 255,
    ThresholdTypes.Binary);

Pixels greater than the threshold become 255 and the remaining pixels become zero. Use ThresholdTypes.BinaryInv when dark objects should become foreground.

Select the threshold automatically

Otsu thresholding calculates a global threshold from the input histogram:

using var binary = new Mat();
double selectedThreshold = Cv2.Threshold(
    grayscale,
    binary,
    thresh: 0,
    maxval: 255,
    ThresholdTypes.Binary | ThresholdTypes.Otsu);

Console.WriteLine($"Selected threshold: {selectedThreshold}");

Cv2.Threshold returns the selected scalar threshold and writes the binary image to the destination Mat. This differs from Python, where both values are normally returned as a tuple.

Handle uneven lighting

Adaptive thresholding calculates a local value for each neighborhood:

using var adaptive = new Mat();
Cv2.AdaptiveThreshold(
    grayscale,
    adaptive,
    maxValue: 255,
    AdaptiveThresholdTypes.GaussianC,
    ThresholdTypes.Binary,
    blockSize: 21,
    c: 5);

blockSize must be an odd value greater than one. Larger neighborhoods follow slower illumination changes; c shifts the local threshold.

Select a color range

Convert BGR input to a color space suited to the selection, then use Cv2.InRange. This example creates a mask for a green hue range:

using var hsv = new Mat();
using var greenMask = new Mat();

Cv2.CvtColor(source, hsv, ColorConversionCodes.BGR2HSV);
Cv2.InRange(
    hsv,
    new Scalar(35, 50, 50),
    new Scalar(85, 255, 255),
    greenMask);

For 8-bit HSV images, OpenCV represents hue in the range 0 through 179. Hue ranges that cross red's wraparound point require two InRange calls followed by Cv2.BitwiseOr.

Remove small mask defects

Morphological opening removes small foreground regions. Closing fills small holes and gaps:

using Mat kernel = Cv2.GetStructuringElement(
    MorphShapes.Ellipse,
    new Size(5, 5));
using var opened = new Mat();
using var cleaned = new Mat();

Cv2.MorphologyEx(
    greenMask,
    opened,
    MorphTypes.Open,
    kernel);
Cv2.MorphologyEx(
    opened,
    cleaned,
    MorphTypes.Close,
    kernel);

Kernel shape and size should reflect the image scale. A large kernel can erase narrow objects or merge nearby regions.

Apply a mask

Pass the mask to Mat.CopyTo to copy only selected pixels:

using var selected = new Mat();
source.CopyTo(selected, cleaned);

The source and mask must have the same width and height. The mask remains single-channel even when the source has three or four channels.

Choose the right operation

  • Use a fixed threshold when lighting and foreground intensity are controlled.
  • Use Otsu or Triangle thresholding when a global histogram separates foreground and background.
  • Use adaptive thresholding when illumination varies across the image.
  • Use InRange when selection depends on an interval or color range.
  • Use morphology to refine a mask after thresholding, not as a substitute for a meaningful segmentation rule.

From C++ or Python

cv::threshold, cv::adaptiveThreshold, cv::inRange, and cv::morphologyEx map to the corresponding PascalCase methods on Cv2. Python examples often combine masks through NumPy operators; use Cv2.BitwiseAnd, Cv2.BitwiseOr, and Cv2.BitwiseNot when the operands are OpenCvSharp Mat objects.

Official OpenCV references

Edit this page