#include <djv_segmenter.h>
Public Attributes | |
| size_t | blockSize |
| Processing block size in pixels. | |
| size_t | maxSizeOfNoiseObj |
| Maximum size of the noise shapes in pixels. | |
| size_t | noiseHueThreshold |
| Noise Hue difference threshold. | |
| size_t | upsample |
| Upsampling for segmentation. | |
| size_t | blurRad |
| Segmenter pre-filter radius. | |
| bool | detectEdge |
| Determines whether to detect edges or not. | |
| bool | thickenShapes |
| Determines whether to thicken the objects or not. | |
| bool | useFgIfEmpty |
| Determines whether to use foreground (0) for empty mask or not. | |
This structure is used to configure the parameters of segmentation.
Processing block size in pixels.
This value should be larger than 4.
The default is 16.
Segmenter pre-filter radius.
This value is segmenter pre-filter radius to reduce noises.
Determines whether to detect edges or not.
This value is to control the detection of edges surrounding a objects. If any objects are connected to an important object which is considered to be vital to the document readability, the objects may reduce the quality of the important object; they changes the color of the important object during the JB2 color processing.
With this flag set to true, you can remove such undesired interference by removing such objects.
The default is false.
Maximum size of the noise shapes in pixels.
Objects larger than this never be noises.
This value should be in [0 64]. If this value is 0, no object is regarded as a noise.
The default is 4.
Noise Hue difference threshold.
This value is the threshold which is used to determine whether an object is a noise or not. If the difference between the hue average of the object and the one of the object background is no more than this value, the object is regarded as a noise and removed.
This value should be in [0 256].
The default is 8.
Determines whether to thicken the objects or not.
This value is to control the thicken-shape process, which improves the readability when used with Color JB2 images.
The default is false.
Upsampling for segmentation.
This value is to upsample the input image for better segmentation result. The output size is not affected by the value.
The value should be in [1 16].
The default is 1.
Determines whether to use foreground (0) for empty mask or not.
This value is to control which layer to choose if the input image has only one color and no need to segment the image. If this value is true, segmenter writes 0 for all the mask pixels; otherwise writes 1 for all the mask pixels.