RECOGNITION SCIENCE, INC. (RSI) has generated a set of macro utilities for Aphelion users. These utilities are designed for direct application to user selected imagery and are not hardwired demos. Some utilities permit easy general purpose image enhancement, some provide support for usage of some of the more complex feature extraction available in Aphelion, and others support the use of feature evaluation and statistical classification capabilities of the Recognition Toolkit. The Recognition Toolkit is an add-on option to the Aphelion system developed by Recognition Science Inc. that includes, feature evaluation, clustering, learning, and classification capabilities. RSI also provides consulting support and can be reached at rsimail@recognitionscience.com. 

   The macros are a free utility set which can be downloaded here. These macros support image enhancement, multi-image combination for noise reduction, sub-image database creation, classification feature generation, and feature classification potential evaluation. (The feature evaluation process requires the RSI Recognition Toolkit add-on to Aphelion.) A user’s guide is included with the RSI Utilities to assist in easy acquaintance with the use on these macros. A run-through testing process is included for the more advanced macros to clarify their usage. The macros are parameterized to permit ready adjustment to new images and new datasets.

    Some results from the image enhancement macros are presented below. The single image enhancement is supported with two key macros. One  performs a clip and stretch operation which expands the dynamic range of images that are not making good use of the full byte range of the image data channels. The second macro provides for edge enhancement using a modified unsharp masking approach. These enhancements can be applied with varying strengths and they can be combined when appropriate.

    In the figure below, four images of a blood sample are displayed. These images show some of the enhancement capabilities in the RSI Utilities. The upper left contains the original image. The upper right is an edge-enhanced version. The lower left image has been clipped and stretched. The image on the lower right was first edge enhanced followed by clipping and stretching. The macro parameters were set to achieve a vivid effect.

Click for enlarged image.

   The multi-image noise reduction macro can be used to merge images that highly overlap but are not necessarily registered at the pixel level. This can be useful when vibration or small sensor movement has effected the data collection. The image below demonstrates the noise reduction of the merge macro using only two input images. The merged picture of part of a fly is shown on the right and has visibly less noise than the two input images.

   Other macros permit the quick creation of large sample sets of sub-image chips that might hold different terrain types, or material types or objects of interest. A macro is provided for extraction of features and assigning class labels to the different chips so that they can be used to develop and train image-based classifiers. An additional macro is included that permits evaluation of and ranking of image features. These macros can ease the development and prototyping of classification algorithms with Aphelion and the RSI Recognition Toolkit. 

   This free macro set is provided as is without any warrantee on performance, but feedback suggestions and comments are very welcome.  You may wish to bookmark this page to easily check back for updates.
 

Download RSIUtilities.exe
 

To find out more about Aphelion in Europe and Asia, click on www.adcis.net
and in all other countries, click on www.aai.com.