KT-Tech Incorporated

KT-TECH'S MORPHOLOGICAL (SHAPE-BASED) IMAGE ENHANCEMENT AND FEATURE EXTRACTION TECHNOLOGY

The ultimate aim in a large number of image processing applications is to extract important features from image data, from which a description, interpretation, or understanding of the scene can be obtained. These features can be edges and boundaries, shape features, spatial features, transform features, etc.

A very well suited approach for extracting significant features from images is morphological (shape-based) processing. Morphology is the study of forms. Morphological processing refers to certain operations where an object is hit with structuring elements and thereby reduced to a more revealing shape. These structuring elements are shape primitives which are developed to represent some aspect of the information or the noise. By applying these structuring elements to the data using different algebraic combinations, one performs morphological transformations on the data. With the addition of some syntactic representations, this shape decomposition can proceed to form a unique representation of the data. In addition, structural elements can be embedded into the syntax to eliminate known noise sources.

Morphological operators are algebraic in form. KT-Tech has developed a higher order grey-level morphological system based on a unique polynomial representation of the grey-level and color imagery. This approach is highly advantageous since polynomials can be manipulated easily, are easily understood, and adaptable for implementation on parallel architectures.

Classical morphological operators are binary and defined for two dimensions. KT-Tech's approach is well suited for optimally addressing all three dimensions of an underlying data set including the intensity. An inherent computational advantage of this approach is that all operations are integer-based. Also, due to the optimal structure of these operators, their computational complexity scales almost linearly with data size, offering a distinct computational advantage.

Other major advantages of this family of operators include effectiveness for low contrast images, modifiability to suit specific problems, computational efficiency, and simplicity.

KT-Tech's morphological image enhancement and feature extraction technology has been utilized for a wide variety of applications.

KT-Tech's morphological and algebraic image operators have been applied to NASA's multi-band LANDSAT imagery to extract and classify relevant features such as land/water boundaries, crop delineation, and to enhance low contrast satellite imagery particularly for cloud identification and ice flow analysis. Figure 1 illustrates the original LANDSAT image of New Orleans. Figure 2 shows the application of a standard edge detector, the Sobel operator, to the original New Orleans image. The application of KT-Tech's morphological image operator to the original New Orleans image is presented in Figure 3. As can be seen, KT-Tech's image operator yields continuous lines along the edges which are one pixel wide whereas the Sobel operator produces disconnected lines which are two or three pixel wide. Note the substantially increased detail obtained through the use of KT-Tech's image operator.

KT-Tech has developed special purpose image processing algorithms to provide grey-level enhancement and edge detection to overcome the loss of visual cues caused by artificial illumination and stark lighting contrast in the area of a robotic arm for NASA applications (Figure 4). The enhancements are designed to improve operator performance of tele-robotic tasks, as in distinguishing the grappling element from the body of a satellite (Figure 5).

KT-Tech's image enhancement technology has also been used to assist in inspecting the integrity of welds for process quality control and safety monitoring. KT-Tech's algorithms enhance images from the welding cup provided by cameras viewing through the torch (Figure 6). Boundary segmentation image operators provide metal flowing and solidification patterns that reveal distinctly when homogenizing is occurring at a welding site and when is not (Figure 7).

In addition, KT-Tech has developed enhancement and feature extraction operators for application to high resolution images from anatomical photographs. KT-Tech has applied its family of morphological and algebraic operators to medical imagery produced by the University of Colorado on their photographs of submilimeter cross-sectional cadaver slices (Figure 8), in support of the National Institutes of Health's "3-D Visible Human" project. KT-Tech's image processing techniques successfully enhanced, delineated, and extracted areas of bone, muscle tissue, fascia, veins, air cavities, and tissue convolutions (Figure 9).

KT-Tech's grey-level image enhancement and edge detection algorithms have been used by the US Postal Service to provide improved semi-autonomous and autonomous resolution of problems of connectivity, segmentation, and image degradations in postal addresses. KT-Tech has developed specific image enhancement operators to improve the performance of character recognition systems.

KT-Tech has also developed an adaptive 3-D world modeler that provides a dynamic approach to continuously tracking a set of known objects in a robotic environment autonomously. This 3-D wire frame modeler generates 2-D synthetic images with control points that vary with perspective. A model matcher compares the synthetic images to the acquired one to provide continuous tracking information for an object, including location, velocity, and pose.

KT-Tech's morphological image enhancement and feature extraction technology can be applied to a wide spectrum of problems including: