Class barPretracedSlideRenderer

Render provided contour slide with given resolution and put rendered image as a first element of self.__imageCache.

It is assumed that self.__imageCache == [].

Apply treshold function for given pixel value. All pixels which are non-white pixels becomes self._tracingConf['GrowDefaultBoundaryColor'] pixels. This procedure assures that output image has only two colours (white and boundary) which is very convinient in further processing.

Parameters:

• pixelColorValue (int) - value of given pixel

Returns: int

255 if white pixel is given, self._tracingConf['GrowDefaultBoundaryColor'] otherwise

Create image cache (self.__imageCache). First cached image is an original image. All other images (up to cacheLevel) are images with succesive grows (applications of MinFilter).

Create outline of the whole brain. The outline serves further as a reference for detecting missplaced labels and obviously as structure representing whole brain.

Outline is created by substracting result of flooding each vBrain label and then inversing the result and tracing resulting bitmap.

Returns: [barPath, ...]

paths representing outline of the whole brain

Dectect labels that cannot be traced as they are missplaced: placed over the contours or outside the brain area.

Returns: [barRegularLabel, ...]

labels that has to be excluded from tracing

Create list of paths basing on the list of labels to be traced (self.__labelsLeft).

Returns: [barPath, ...]

paths created by tracing individual labels

Create list of paths and corresponding labels by tracing areas that were not labeled in pretraced files (unlabeled areas). This process is rather complicated so few words of explanation:

1. We look for patches of N or more pixels. Only such areas are considered as unlabelled areas. Smaller areas are most probably residual white pixels and should be ommited.
2. Every group of two or more adjacent white pixels is flooded on "1".
3. Patch of N or more white pixels is flooded with value "2" after tracing.

After fiding unlabelled areas we can find 4 values of pixels:

1. "0": areas outside brain
2. "1": pixels of more than one adjacent pixels but less than N
3. "2": patches of N or more white pixels

Returns: ([barPath, ...], [barRegularLabel, ...])

pair of lists of paths and corresponding labels generated during detections of unlabelled areas.

Clear all temporary class attributes created for tracing purposes. Invoked by self.trace method as the last step of tracing.

Extract a path defined by coordinates of the given seed label.

If grow level for given seed label is not provided, perform automatic grow level selection, and finally create path with all properties.

Parameters:

• seedLabel (barRegularLabel) - seed label

Returns: [barPath, ...]

list of paths resulting from tracing of the given source image and provided seed label.

Perform all operation releted with tracing bitmap into list of paths:

1. create DOM structure from PoTrace SVG output,
2. create single path segments instead of long bezier paths,
3. convert path coordinates to absolute,
4. reduce transformation rules (using svgfix module.

Parameters:

• sourceImage (PIL.Image.Image) - image for tracing
• seedLabel (barRegularLabel) - seed label
• invert (bool) - determine if image has to be inverted before tracing

Returns: [barPath, ...]

paths resulting from tracing of the given source image and the provided seed label.

Create barPath from provided SVG path element (result of tracing) and seed label. Resulting path has proper identifier and structure name, however it has black colour assigned.

Parameters:

• pathElem (SVG path element) - xml.dom.minidom.Node
• seedLabel (seed label) - barRegularLabel

Returns: barPath

path representation created from provided SVG path element

Generate proper path identifier of the path related to the seed label.

Parameters:

• seedLabel (seed label) - barRegularLabel

Returns: str

identifier of new path generated using self._tracingConf['NewPathIdTemplate']

Mark traced areas on __brainOutline. Depending on source marking is performed in different way. Region denoted by vBrain regions are changed to black (0) while areas from tracing normal labels are changed to 100.

Parameters:

• imageToSubstract (PIL.Image.Image) - PIL image
• source (str) - determines source of imageToSubstract (may be either 'Struct' or 'Brain').

Apply flood fill to given image using coordinates passed in the seed label.

Description:

1. Extract coordinates and structure name from seedLabel.
2. Rescale SVG coordinates to rendered image coordinates
3. Fill given image with black color
4. Extend filled region by applying MinFilter filter. Do it l+1 times
5. Return flooded image or return number of black pixels depending on retNPIX value.

Parameters:

• sourceImage (PIL.Image.Image) - image to be floodfilled
• seedLabel (barRegularLabel) - seed label
• retNPIX (bool) - if True, floodfilled image is returned along with number of replaced pixels; otherwise only flooded image is returned

Returns: (PIL.Image.Image, int) or PIL.Image.Image

flooded image (and number of flooded pixels if requested)

Determine if bitmap with given seed label is suitable for tracing. There are briefly two conditions which need to be satisfied in order to allow the tracing:

1. Seed has to be placed within brain contour. The brain contour is defined at the begining of tracing procedure.
2. Seed has to be placed at white pixel. If seed points into non-white pixel is it a) border between two structures; b) area outside brain contour.

In both cases debug information is printed.

Parameters:

• im (PIL.Image.Image) - image to be floodfilled
• seedLabel (seed label) - barRegularLabel

Returns: bool

True if image is suitable for tracing (according to testing conditions), False otherwise

Find white pixels in the image. If found pixel has white neighbourhood, its coordinates are returned, otherwise iteration continues until all white pixels are spotted.

Current algorithm is rather ineffective. Main reason is that function has to exit if patch of white pixels is spotted. Then function is invoked again and iteration starts from beginning which is extremely ineffective. What should be done in such case? Perhaps implementation should include iterator. The problem is that we need to change array while we iterate over this array.

Returns: (int, int)

(x, y) image coordinates of first spotted pixel that belongs to patch (two or more adjecent pixeles) of white pixels; if there is no such pixel, None

Determine if label is placed outside brain outline according to values of self.__brainOutline.

Parameters:

• seedLabel (seed label) - barRegularLabel

Returns: bool

True if label is placed outside brain, False otherwise

Determine if label is not placed over boundary.

Parameters:

• im (PIL.Image.Image) - image to verify seed label location
• seedLabel (seed label) - barRegularLabel

Returns: bool

True if label is not placed over contour, False otherwise

Save provided image.

Parameters:

• im (PIL.Image.Image) - bitmap image to be saved
• filename (str) - name of the output file (only filename - without path; provide .png extension)

Save provided SVG.

Parameters:

• svgdom (xml.dom.minidom.Document) - SVG image to be saved
• filename (str) - name of the output file (only filename - without path; provide .png extension)

For images from self.__imageCache perform floodfilling with seed label.

Parameters:

• seedLabel (barRegularLabel) - seed label

Returns: ([PIL.Image.Image, ...], [int, ...])

pair of lists of floodfilled images and area of floodfill in pixels (both lists have the same length and are corresponding to some prefix of self.__imageCache)

Output warning message to stderr and dump invalid image into a file.

Parameters:

• im (PIL.Image.Image) - image to be dumped
• seedLabel (seed label) - barRegularLabel