CIB image toolbox technical manual
9. Segmenters
RDO_Segmenter
ColorQRSegmenter
SplitChannelsSegmenter
RangesSegmenter
Interface of image segmentation by color ranges.
This segmenter allows to extract from image to separated images all pixels with colors that fall within user defined ranges (or if it is required pixel that do not fall within these ranges). These ranges can be set by definition of lower and upper range boundaries:
<RangesSegmenter GetMasks="0"> <UsedColorSpace>RGB</UsedColorSpace> <LowerBoundaries>0,0,0;130,55,80;219,200,130</LowerBoundaries> <UpperBoundaries>128,128,128;209,190,190;255,255,191</UpperBoundaries>
</RangesSegmenter>
Property name |
Description |
Default value |
UsedColorSpace |
Colorspace of an image |
RGB |
LowerBoundaries |
Lower boundaries of a range |
0,0,0;127,127,127 |
UpperBoundaries |
Upper boundaries of a range |
128,128,128;255,255,255 |
GetMasks |
Flag for masks |
0 |
RDO_Segmenter
Segmenter for extracting layers from an image.
This segmentation procedure is based on article Cheng, H. and Bouman, C.A., 2001. Document compression using rate-distortion optimized segmentation. Journal of Electronic Imaging, 10(2), pp.460-475. But RDO_Segmenter is modification of algorithm from the article.
Parameters of this filter can be described in profile:
<ProcessingProfile>
...
<RDO_Segmenter> <ImageBlockSize>8</ImageBlockSize> <IsTextBinaryMaskRequired>true</IsTextBinaryMaskRequired> <ThreshHigh>150.0</ThreshHigh> <ThreshLow>25.0</ThreshLow> <ThreshMean>200.0</ThreshMean> <DCTCoeff>400</DCTCoeff> <ThreshVariance>25</ThreshVariance> <IsDownsamplingRequired>false</IsDownsamplingRequired> <RatioForeground>4.0</RatioForeground> <RatioBackground>2.0</RatioBackground>
</RDO_Segmenter>
...
</ProcessingProfile>
Property name |
Description |
Default value |
ImageBlockSize |
Size of image blocks for processing |
8 |
IsTextBinaryMaskRequired |
Flag for result layers filling |
true |
ThreshHigh |
High threshold for the segmenter |
150.0 |
ThreshLow |
Low threshold for the segmenter |
25.0 |
ThreshMean |
Mean threshold for the segmenter |
200.0 |
DCTCoeff |
Threshold for discrete cosine transform |
400 |
ThreshVariance |
Variance threshold |
25 |
IsDownsamplingRequired |
Flag for optional background and foreground downsampling |
false |
RatioForeground |
Scale factor for foreground downsampling |
4.0 |
RatioBackground |
Scale factor for background downsampling |
2.0 |
ColorQRSegmenter
Segmenter for extracting layers from a QR RGB image (experimental). It only accepts 1 input image, and outputs 3 images, one per channel.
Parameters of this filter can be described in profile:
<ProcessingProfile>
...
<ColorQRSegmenter> <IsTextBinaryMaskRequired>true</IsTextBinaryMaskRequired>
</ColorQRSegmenter>
...
</ProcessingProfile>
Property name |
Description |
Default value |
MaxSide |
Longest side for rescaling |
0 |
CleanBackground |
If enabled it attempts to clean the background |
false |
Blur |
If enabled it will execute blurring |
false |
Sharp |
If enabled it will add a sharp filter |
false |
HueDiscretization |
If enabled it will use HUE discretization to process the image |
false |
Grayscale |
If enabled, final image will not be binarized |
false |
SplitChannelsSegmenter
This segmenter allows to separate multichannel image to few grayscale images with one channel.
There are no parameters for this segmenter.