This proposal proposes the synergistic integration of several methods for mining images, detecting, correlating and evaluating the existence of artifacts due to either hidden information or changes or target patterns or noise. The first method is based on the Pixels Flow Functions (PFF) able to detect changes in images by projecting the pixel values vertically, horizontally and diagonally. These projections create "functions" related with the average values of pixels summarized horizontally, vertically and diagonally. These functions represent image signatures. The comparison of image signatures defines differences among in images. On the changes discovered by the PFFs morphological image operations will be used for mining the differences. The second method is based on a heuristic graph model, known as Local-Global Graph (LGG), for evaluating modifications in digital images and defining patterns and determining structural associations (relationships). The LGG is based on segmentation and comparing the segments while thresholding the differences in their attributes. The third method is based on stochastic Petri-net graphs (SPNG) able to detect and describe functional relationships (formations) among the changes and patterns and provide first stage interpretation (or knowledge discovery). The next part of the methodology proposed here is the fusion of multimodal representation (visual, IR, thermal , radar) of images for more accurate detection and extraction of the right target patterns. The last part of the research approach here is the tracking and extraction of target patterns from sequences of images. First stage results of each of the first and second methods, implemented in C++, are presented as a first level proof of concept regarding the feasibility of the proposed work.
Benefits: The anticipated benefits from this project are tool-methodologies for: 1. Mining images and sequence of images for detecting similarities and differences 2. Detecting patterns from images and sequence of images 3. Determining time associations and formations of patterns and their relationships 4. Fusing multimodal representation of images 5. Tracking and extracting targeted patterns from sequences of images The commercial applications of the outcome of this STTR effort are: 1. Document processing 2. Handwritten recognition 3. Image understanding 4. Video Analysis 5. Biometrics based Security 6. Face Recognition 7. Biomedical Imaging 8. Biological Imaging 9. Surveillance Systems 10. etc. Abstract: Mining, Similarities, Images, Sequences of images, Detection, Differences, Target patterns
