The ROI can be managed by the buttons

• Paste - paste the ROI from the image to the scorpion clipboard
• Copy - copy the ROI to the image from the scorpion clipboard

Point & Click Clipboard Support

The rectangular ROI is defined by four points.

 One point will change the center point.

Colors

• Selected Reference - displays the image of the selected reference

• - Adds a color with a given name
• - Removes a color with it's reference images
• - Adds the current image as a reference to the selected reference
• - Removes the selected reference

Colors Mouse Menu

• Add color

• Active - toggle the active flag (inactive colors are not reported as results, and are listed with yellow checkmark in front of the name)

• Set code - a colors numeric value

• Set display color - for visualisation

• Rename color

• Delete color

• Delete all references

Reference Mouse Menu

• Add reference

• Rename reference

• Delete reference

• Delete all references

Classification

Classification type -

Classification of each rectangle within the grid can be done using color, texture or a combination:

• Color only - disable texture classification

• Texture only - disable color classification

• Color, then texture - color is used to create regions in the image; texture information is added using a voting

• Texture, then color - texture is used to create regions in the image; color information is added using a voting

• Combined - regions are created with both similar color and texture

Color features

Color distance

• Activates the color planes to be part of the distance calculation

  • H - activates Hue

  • S - activates Saturation

  • I - activates Intensity

  • H std - activates Standard Deviation Hue

  • S std - activates Standard Deviation Saturation

  • I std - activates Standard Deviation Intensity

• Weighting factors - coeffisients for the feature distance formula

  • H - Mean Hue - default 4 - increases the importance of the Hue value

  • S - Mean Saturation - default 1

  • I - Mean Intensity - default 1

  • H std - Standard Deviation Hue

  • S std - Standard Deviation Saturation

  • I std - Standard Deviation Intensity

• Weighing of H by

  • S - Saturation - default on

  • I - Intensity - default  on

• Feature Distance Formula

  • H weighted by square  root of normalized S and I as selected - recommended Feature Distance

  • H linearly weighted by normalized S and I as selected

  • Normal measure H, S and I equivalent - d = sqrt(H*H + S*S + I*I)

Note: adding the standard deviation coeffisient to the distance measure makes it possible to separate non-monochrome colors.

Color constraints

• Maximum distance - the maximum distance to the current image for a reference image to be accepted

• Minimum separation - the minimum separation between the best and the second best color for the color to be accepted

Note: The separation is often used to get only clear colors matched - A typical value can be 6.

Texture features

Image decimation

• Active - activates image decimation

  • Horizontal - horizontal decimation factor

  • Vertical - vertical decimation factor

  Note - Image decimation speeds up the TextureMatcher - it is by default off

Normalizing - used to make the texture matcher less sensitive to intensity variations

• Method

  • 1-Offset normalizing - default

  • 2-Offset- and mean-gain normalizing

  • 3-Offset- and std-gain normalizing

Texture distance - specifies how to calculate the distance - difference is the best method

• Method

  • 1-Dissimilarity - default

  • 2-Symmetric dissimilarity

  • 3-Similarity without weights

  • 4-Similarity

• Active coefficients

  • By default all coeffisient except LL and HH are included - to deviate from default one verify the performance on a large image set

Texture constraints - to specify an absolute criteria to accept a texture

• Maximum distance - the maximum distance to the current image for a reference image to be accepted

• Minimum separation - the minimum separation between the best and the second best texture for the texture to be accepted

Separation

To be completed

This is a list of statistics for the defined colors. Within each color group the least similar sample is found, and its mean distance to the other references in the same group is calculated (Max int dist). Further, this reference's mean distance to each of the other color groups is found, and the nearest result is reported (Closest col and Min ext dist).

The ratio between Min ext. dist and Max int. dist in dB is reported as mid. A negative value here means the reference is on average closer to another group than its own. 

Finally, the ratio between the least similar sample's closest match to any reference in other color groups and the closest match within its own group is reported, also in dB. Even if the mid value is low, this min value can be positive, which means the reference is still closer to its own group, though not on average.

These results can be used to spot outliers in a color group. If the mid separation in dB falls below the Mid color separation warning threshold, and the min separation below the Closest separation warning threshold, the checkmark in front of the color name turns red. You can spot the outlier reference under the Setup page as well, where both the color and the reference will be marked in red.

NOTE: this has NO effect on the tool's results, it is just an aid in setting up well defined color sets.

Blobs

Blob analysis

Morphology

A blob analysis is performed on the classified rectangles, creating connected areas. The blobs can be grown and shrunk using a Morphology string.

Hole filter

Blob holes can be ignored based on size.

Results sorted by

A number of "the best" blobs are reported in the Results pane; here you decide which blobs are considered "best".

Result string

• Create code grid string result: create a python 2d tuple with classification codes (color+texture codes)

  • Raw - result from raw blob analysis (prior to morphology and constraints)

  • Final - result after morph/constraints. If blob collision, code is overwritten

Constraints

The blobs resulting from the analysis can be filtered out by a set of criteria:

• Contour area - the total area of the blob

• Blob area - contour area minus area of any holes

• Hole count - number of holes in the blob

• Axis ratio - length/width of bounding box

• Angle - angle of bounding box major axis

• Length of bounding box

• Width of bounding box

• Blob circularity - closeness to circular shape (max 1). Smoothing blobs with a higher factor will help

The bounding box is the smallest possible rectangle containing the blob.