VLSI design and implementation of a real-time image segmentation processor

Bhanu, Bir; Hutchings, Brad; Smith, K.F.

doi:10.1007/bf01211450

Cited by 7 publications

(5 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For an image of size 300 200, it takes several minutes on a SunSparc10 to finish the decomposition. We are investigating very large-scale integration implementation [5] of Gabor filters, which can dramatically reduce the cost of Gabor image decomposition. Since the models are generated off-line and stored as Gabor grid, the complexity of matching is significantly reduced by our speed-up grid placement and flexible matching algorithm.…”

Section: Discussionmentioning

confidence: 99%

Gabor wavelet representation for 3-D object recognition

Bhanu

1997

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

This paper presents a model-based object recognition approach that uses a Gabor wavelet representation. The key idea is to use magnitude, phase, and frequency measures of the Gabor wavelet representation in an innovative flexible matching approach that can provide robust recognition. The Gabor grid, a topology-preserving map, efficiently encodes both signal energy and structural information of an object in a sparse multiresolution representation. The Gabor grid subsamples the Gabor wavelet decomposition of an object model and is deformed to allow the indexed object model match with similar representation obtained using image data. Flexible matching between the model and the image minimizes a cost function based on local similarity and geometric distortion of the Gabor grid. Grid erosion and repairing is performed whenever a collapsed grid, due to object occlusion, is detected. The results on infrared imagery are presented, where objects undergo rotation, translation, scale, occlusion, and aspect variations under changing environmental conditions.

show abstract

Section: Discussionmentioning

confidence: 99%

Gabor wavelet representation for 3-D object recognition

Bhanu

1997

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

show abstract

“…• The adaptive image segmentation system may soon be able to benefit from advances in parallel computing and VLSI technologywhich are now beginning to produce chips that can perform the image segmentation process in real time [8]. These hardware improvements would make it possible to achieve high quality image segmentation results at near-real-time processing rates.…”

Section: Discussionmentioning

confidence: 99%

“…They include edge detection [20], region splitting [49,50,53], region merging [36,55], clustering [19,46], surface fitting [55], rule-based expert systems [47], relaxation [4,6,8,9,55], and integrated techniques [27,35,39,55]. In this chapter, we first briefly discuss techniques based on edge detection, and region splitting and region growing, and then present the details of the Phoenix image segmentation algorithm [41,61] that has been used in this research.…”

Section: Techniquesmentioning

confidence: 99%

Genetic Learning for Adaptive Image Segmentation

Bhanu¹,

Lee²,

Kanade³

1994

Self Cite

View full text Add to dashboard Cite

Genetic learning for adaptive image segmentation / Bir Bhanu, Sungkee Lee. p. cm. --(The Kluwer international series in engineering and computer science ; 287. Robotics) Includes bibliographical references and index. ISBN 978-1-4613-6198-5 ISBN 978-1-4615-2774-9 (eBook) 10 SUMMARY REFERENCES INDEX vii 218 220 255 261 269 Preface xixThis book also explores a hybrid search scheme that combines genetic algorithms and hill climbing for adaptive image segmentation. It provides experimental results and compares its performance and efficiency with that of the baseline approach that uses only the genetic algorithm.The book further develops the baseline adaptive image segmentation system for multiobjective optimization. The global and local quality measures are optimized simultaneously for adaptive image segmentation. Experimental results are provided and compared with the baseline approach.The adaptive segmentation system presented in this book is very fundamental in nature and is not dependent on any specific segmentation algorithm or sensor data (visible, infrared, laser, etc.).Zucker [70] summarizes the above conditions as follows. The first condition implies that the segmentation is complete: every pixel is in a region. This means that the segmentation algorithm should not terminate until every pixel in an image is processed. The second condition requires that regions are connected, i.e., regions are composed of contiguous pixels. The third condition determines what kind of properties the segmented regions should have. It is either a syntactic image characteristic (e.g., intensity, color, texture) or corresponds to some semantic interpretation. The fourth condition expresses the maximality of each region in the segmentation.This list of conditions for an image segmentation does not lead to a unique segmentation algorithm, but does suggest important aspects of such algorithms. CHARACTERISTICS OF THE IMAGE SEGMENTATION PROBLEMImage segmentation is typically the first and most diflicult task of any automated image understanding process. It is also known as an ill-defined problem. It refers to the grouping of parts of an image that have "similar" image characterist(cs. All subsequent interpretation tasks, including o~iect detection, feature extraction, object recognition, and classification, rely heavily on the quality of the segmentation results.Despite the large number of segmentation techniques presently available [5,27,35], no general methods have been found that perform adequately across a diverse set of images. There are many reasons why we do not have a general-purpose image segmentation system that will work for all images:First, a two-dimensional image represents potentially an infinite number of scenes. For instance, an image of 128 by 128 pixels in size and 8 bits in pixel resolution can represent, (2 8 )128xI2R == 1039457 different scenes. For such a variety of scenes, it is impossible to define a single logical predicate that produces good segmentation for each possible image.Second, images of natural scenes...

show abstract

“…The segmentation process is an important early stage in image analysis and pattern recognition; for example, a well-chosen threshold value is of paramount importance to a binary-level pattern recognition system. Bhanu et al (1990) described a VLSI design to implement a real-time segmentation processor. However, a real-time implementation of a segmentation algorithm is quite complex and each algorithm may need dedicated hardware.…”

Section: Introductionmentioning

confidence: 99%

Real-time image segmentation system for IC die patterns

Lee

1995

International Journal of Electronics

View full text Add to dashboard Cite

A real-time image segmentation system realized for extracting the bond pads in IC die images is presented. The system can work in current IC manufacturing machines with binary-level pattern recognition systems under nearly uniform lighting conditions. The system uses a general-purpose digital signal processor DSP56001 to realize the global thresholding algorithm, and most of the functional blocks are implemented using programmable logic devices. The selection of the thresholding algorithm is based on the characteristics of industrial IC die images. IntroductionImage segmentation is a process which divides images into several meaningful regions so that the segmented image can be used for further analysis or interpretation. The segmentation process is an important early stage in image analysis and pattern recognition; for example, a well-chosen threshold value is of paramount importance to a binary-level pattern recognition system. Bhanu et al. (1990) described a VLSI design to implement a real-time segmentation processor. However, a real-time implementation of a segmentation algorithm is quite complex and each algorithm may need dedicated hardware. As a result, redesign of a segmentation processor is required when a new segmentation algorithm is proposed. On the other hand, if the segmentation algorithm is implemented using a digital signal processor (DSP), then the realization of a new algorithm may only require modification of software.In the IC manufacturing industry, the advances in application-specific and semicustom ICs has led to the evolution of high lead-count and high-density semiconductor packages. Hence, the bond pads have become smaller and smaller and the lead frame narrower and narrower. This provides challenges for the wire bonding technology. Shah et al. (1988) identified several new technological requirements in the design of the next-generation of wire bonders. These include high placement accuracy and minimum set-up time. These requirements are partly related to the pattern recognition system for the bonding machines, where the pattern recognition system has to locate the bond pads or bonding leads quickly and accurately.Since modern wire bondedrs are capable of bonding more than ten wires per second, in order not to degrade the machine throughput a fast and reliable pattern recognition system is urgently needed. Binary-level pattern recognition systems best suit this requirement as these systems can recognize a pattern from a binary-level

show abstract

VLSI design and implementation of a real-time image segmentation processor

Cited by 7 publications

References 14 publications

Gabor wavelet representation for 3-D object recognition

Gabor wavelet representation for 3-D object recognition

Genetic Learning for Adaptive Image Segmentation

Real-time image segmentation system for IC die patterns

Contact Info

Product

Resources

About