5.4 Extension of Functionality

It is also planed to extend the laboratory by a set of image recognition functions. This will be implemented by students of San Diego State as part of class assignments and projects. This is a rough summary of functions that are planned:

• Object Separation Through Thresholding: This function separates objects from the background within an image by partitioning of the gray scale histogram of the image according to single or multiple thresholds.

• Region Growing by Pixel Aggregation: This function collects pixels within an image into larger regions that become candidates for individual objects. It starts from a set of seed pixels that are automatically generated or selected by the user. The result is a set of regions of neighboring (connected) pixels that have similar properties, e.g. the difference between their gray scale values is smaller than a predefined threshold.

• Region Splitting and Growing: This function identifies regions within an image, by first subdividing the image into arbitrary disjoint regions and then merging those regions that have same attributes into larger areas.

• Patter Matching by Correlation: The goal of this method is to identify an object (i.e. sub-image) within another larger image by using the correlation between these two images.

• Pattern Matching within the Frequency Domain using FFT: This approach is often more efficient than the previous method, i.e. Pattern Matching by Correlation. This is especially the case for larger objects (sub-images).

• Hough Transform: This method is used to identify figures (lines, circles...) within an image. This method is tolerant of gaps in the figures and is relatively unaffected by the noise within the image.