Detailed explanation of contour extraction parameters in vision software

Threshold Parameters – Parameter Details

In the visual contour edge extraction function, threshold parameters are used to control the sensitivity of edge detection. The main points are as follows:

1. The threshold parameter typically ranges from 0 to 255. A higher value results in higher detection sensitivity, while a lower value results in lower sensitivity.

2. The principle for selecting the threshold parameter is to preserve the target contour while eliminating as much noise and irregular edges as possible. Therefore, the appropriate threshold should be chosen based on the quality of the input image.

3. For high-quality images, a higher threshold (e.g., above 200) can be selected to detect more detailed edges. For images with significant noise, a lower threshold (e.g., below 150) should be used to avoid detecting excessive noise points.

4. If the threshold is set too low, contour edges may become blurred and lose detail. If set too high, a large number of noise points may be detected, resulting in over-segmentation.

5. Different algorithms and implementations have varying sensitivity to threshold parameters. Therefore, threshold selection should also take into account the specific algorithm being used.

6. The best practice for threshold selection is to test multiple threshold values on a training dataset, evaluate the visual quality and quantitative metrics of the results, and choose the threshold that achieves both satisfactory visual performance and optimal metrics. It is also necessary to consider variations in input data and dynamically adjust the threshold accordingly, which requires some accumulated experience.

7. In practical applications, computer vision frameworks often provide automatic threshold optimization functions. These can automatically determine optimal threshold parameters on training data, simplifying the parameter tuning process for users. The effectiveness of such automatic optimization improves as the training data becomes more abundant and the model is further refined.

Expansion Parameters – Parameter Details

In the visual contour edge extraction function, expansion parameters are used to control edge width. The main points are as follows:

1. The expansion parameter typically ranges from 0 to 20. A higher value produces a more pronounced expansion effect, while a lower value produces a less pronounced effect.

2. The principle for selecting the expansion parameter is to achieve the desired edge width while maintaining correct contour edges. In general, expansion values between 3 and 10 are sufficient for most application scenarios.

3. If the expansion value is set too low, edges may appear overly thin and sharp. If set too high, edges may appear overly wide and blurred. Both cases can adversely affect subsequent processing.

4. The selection of expansion parameters should also consider the resolution of the input image. For high-resolution images, the same line width in pixels may produce different visual effects. Therefore, smaller expansion values are generally recommended for high-resolution images.

5. Different algorithms and implementations have varying sensitivity to expansion parameters. The Canny algorithm is more sensitive to expansion parameters, while other algorithms such as LoG (Laplacian of Gaussian) are relatively less sensitive. Therefore, expansion parameter selection should also take the specific algorithm into account.

6. Expansion parameter selection should similarly be tested on training data, evaluating visual effects under different parameter values to select the optimal one. In addition, multiple versions can be customized based on business needs, such as fine-edge, standard, and thick-edge versions, to accommodate different application scenarios. In practice, computer vision frameworks also provide automatic expansion optimization features that can intelligently recommend better expansion parameters based on image content, thereby simplifying the tuning process and improving efficiency.

Contour Level Parameters – Parameter Details

In the visual contour edge extraction function, contour level parameters are used to control the number of detected edges. The main options are as follows:

1. All Levels: Detects all possible edges in the image, including subtle edges. The range is typically 1 to 3, with higher values detecting more edges. This is suitable for scenarios that require preserving maximum edge information, though the visual result may be more complex.

2. Level 1: Detects only the most prominent contour edges in the image while filtering out a large number of subtle edges. The range is typically 1. This is suitable for primary edge analysis and scenarios requiring visually concise results.

3. Outermost: Detects only the outermost closed contour edges at the periphery of the image. The range is 1. This is mainly used for object shape detection and segmentation scenarios.

4. Innermost: Detects only inner-layer contour edges while filtering out outer edges. The range is -1 to -3, with larger absolute values detecting fewer inner-layer edges. This is used for scenarios involving the detection and analysis of internal object structures.

Fitting Method Parameters – Parameter Details

In visual contour edge extraction, fitting method parameters control the smoothness of edge curves. The main options are as follows:

1. Sharp Corner Priority
Preserves sharp-angle and pointed features without smoothing.

(1) Angle Type Classification:

• Convex: Prioritizes convex corner points (angle < 180°). Suitable for sharp-angled contours like faces or buildings.

• Concave Only: Preserves only concave corners (angle > 180°), filtering out convex ones. Suitable for rounded contours like cars or aircraft.

• Ignore Convex/Concave: Retains all corner points regardless of convexity/concavity. Suitable for various contour types, though accuracy may be reduced.

(2) Angle Threshold:
Sets the minimum angle to qualify as a sharp corner. Only angles exceeding this threshold are considered sharp. Range: 20°–90°. Smaller thresholds identify more sharp corners. For straight-edged contours like buildings, use a higher threshold (e.g., 60°–90°) to avoid misclassification.

(3) Angle Noise Reduction:
Filters out noise and fine details based on surrounding angle information. A point is valid only if its angle exceeds the local average plus a noise reduction value. For detailed/noisy contours like faces, use a lower value (e.g., 5°–10°) to improve accuracy.

(4) Use Case:
Best applied when contours contain obvious sharp corners.

2. Curve Fitting
Smooths edges using curve fitting algorithms. This is a commonly used contour processing method that fits contours to a smooth curve, further improving contour accuracy and smoothness.

(1) Use Case:
When contours are relatively smooth, curve fitting produces a smoother fitted contour.

3. Edge Precision
A parameter used to control the edge precision of curve fitting results. This parameter typically defines the distance threshold—points within this distance from the edge are considered edge points, thereby affecting the edge precision of the curve fitting. The edge precision parameter controls the accuracy and smoothness of the fitted curve at the edges. A smaller edge precision value can fit more detailed edge features but may also introduce noise and oscillation. A larger value produces a smoother curve but may lose edge detail information.

The edge precision parameter is typically a floating-point value ranging from 0 to 1, with different values affecting the edge precision and smoothness of the curve fitting. In general, the edge precision parameter should be selected based on specific data characteristics and application scenarios to achieve a balance between edge accuracy and smoothness.

Custom DPI Parameters – Parameter Details

The Custom DPI function allows users to set the DPI (Dots Per Inch) value of the output image to control its size and clarity. DPI is an important parameter for measuring image output clarity.

After contour edge extraction processing, images are often output for subsequent processing or display. Users can control the output image size and clarity by adjusting the DPI value to meet different requirements.

The Custom DPI parameter typically ranges from 1 to 1000, with different values affecting output image size and clarity. A higher DPI value results in a clearer image but also increases the file size. Therefore, the DPI should be selected based on the specific application scenario.

The Custom DPI function is applicable in fields such as image processing, computer vision, and robotic vision. For example, in object detection and tracking, high-clarity target contour images are required for subsequent processing and analysis. In robotic vision, contour images with moderate size and high clarity are needed for real-time control and decision-making.

By using the Custom DPI function, users can flexibly control output image size and clarity to meet various application requirements, thereby improving the effectiveness and accuracy of visual contour edge extraction processing.

Enable Mark Point Extraction Parameters – Parameter Details

This function is used to add marker points during the contour extraction process to facilitate subsequent image processing or analysis. The size of the marker points is specified in millimeters (mm) and can be set to any integer value from 1 to 100. The marker point size determines the visibility and space occupied by the markers. Generally, larger marker points are easier to detect and identify but also occupy more space.