Image Filtering

## Image Filtering

Image filtering is to perform preset mathematical operations on each pixel (or pixel neighborhood) in the image, screen and retain useful information, remove useless interference, and finally output an image that meets subsequent processing requirements

## Module Principle
Only spatial domain filtering is considered, processing is performed in the spatial domain of the image (i.e., the two-dimensional coordinate domain of pixels), and mathematical operations (such as addition, subtraction, multiplication, division, weighted average, sorting, etc.) are performed on each pixel point and its surrounding neighborhood pixels, and the operation result is used to replace the original pixel value.

## Usage Scenarios
During shooting (camera noise), transmission (signal interference), and segmentation (binarization noise), various useless noises (such as salt and pepper noise: black and white spots, Gaussian noise: uniform blur interference) will be generated. In order to remove these noises and make the image smoother and clearer, avoiding noise interference with subsequent feature extraction and recognition, this module can be used.

## Usage Method
For example: Input an original image, set the filter kernel size, and perform mean filtering on it. It can be seen that the edges of the image are blurred.

![](/media/202601/2026-01-16_095211_8343980.9196824439963361.png)

Effect:
![](/media/202601/2026-01-16_095236_1604060.7419640319923508.png)

In the table on the left, you can add, delete, move up, and move down to set the filter type (multiple types can be set). On the right is the parameter setting when selecting different filter types. You can set the mask to select the enabled region (detection mask part) and the inverted region (shield mask part).
![](/media/202601/2026-01-16_095403_8909350.32815044469221055.png)

![](/media/202601/2026-01-16_095817_7039280.837738771188455.png)

### Usage Example
[Record operation video]

## Parameter Configuration Description

![](/media/202601/2026-01-16_095829_4628930.05352428616743432.png)

Dilation Filter: Gray dilation, set kernel size and kernel type, perform gray dilation on pixels; for example, select a rectangular, 5*5 kernel to perform gray dilation on the pixel at the red dot in the figure, and replace the pixel gray at the red dot with the maximum gray value in the surrounding 5*5 box.
![](/media/202601/2026-01-16_100845_1617230.14502148631581713.png)
Erosion Filter: Gray erosion, opposite to gray dilation, select the minimum surrounding gray for replacement.
Gaussian Filter: Weighted average of neighborhood pixel gray values according to Gaussian distribution.
Mean Filter: Take the average value of surrounding pixel gray values for replacement.
Median Filter: Sort the surrounding pixel gray values and take the middle gray value for replacement.
Image Transformation: This does not belong to image filtering, it is an operation such as rotating and mirroring the image.
Edge Extraction: This does not belong to image filtering, set gray threshold for edge detection, as shown in the figure:
![](/media/202601/2026-01-16_104434_5395770.03872416897399045.png)

Gamma Correction: Enhance image, correct image brightness and contrast, match human visual perception characteristics, compensate for non-linear response of image acquisition / display devices. Gamma value γ>1, image becomes darker, Gamma value γ<1, image becomes brighter.
Brightness Adjustment: Perform addition operation on original pixel gray directly in the original image to make the image brighter.
Manual Binarization: Set pixels within the set threshold interval in the image to gray 255, and others to 0, as shown in the figure:
![](/media/202601/2026-01-16_110311_1149210.4659579830005215.png)

Adaptive Binarization: Local binarization for images with uneven illumination, avoiding "one size fits all" global threshold, optional mean weighting and Gaussian weighting methods.
Denoise Binarization: Can filter some regions with insufficient area or excessive area. Similar to manual binarization function.
Binarization Type: Manual White, indicating gray threshold interval is (Gray Threshold-255); Manual Black, indicating gray threshold interval is (0-Gray Threshold); Automatic White, indicating gray threshold interval is (128-255); Automatic Black, indicating gray threshold interval is (0-128); Area Screening: Retain gray values of areas meeting the interval, others set to 0.

![](/media/202601/2026-01-16_113202_9067490.11396885296934511.png)

Region Pixel Filling: A mask must exist to decide where to fill pixels, and the pixels of the selected region are uniformly set to a fixed pixel value.
Linear Gray Stretch: Linearly map the original gray interval of the image to the target gray interval (0-255).
Histogram Equalization: Through non-linear gray value remapping, "stretch" and "uniformly distribute" the originally concentrated histogram to the entire 0~255 gray interval, so that the number of pixels at each gray level in the image is as close as possible, thereby maximizing the gray dynamic range and improving the global contrast.
Image Self-Multiplication: A linear image enhancement, multiplying the pixels in the image by a coefficient to obtain a new image.
Edge Sharpening: Enhance regions with abrupt gray changes in the image and suppress regions with gentle gray changes.
Image Enhancement 1: Enhance high-frequency components (edges, details) of the image to improve image clarity and contrast.
Preserve Boundary Denoising: The core uses mean filtering and retains edge information in the image.
![](/media/202601/2026-01-16_110630_8725950.538748131075354.png)

## Output Description

Running Result: Output True if filtering operation is completed, otherwise output False.
Output Image: Monochrome image, color image is not supported for now.
![](/media/202601/2026-01-16_142816_3349170.8934412868376994.png)

---

Click here to jump to the corresponding Chinese manual page: [Image Filtering](/doc/53)