Farshid Arman

Content-based browsing of video sequences

A novel methodology to represent the contents of a video sequence is presented. The representation is used to allow the user to rapidly view a video sequence in order to find a particular point within the sequence and/or to decide whether the contents of the sequence are relevant to his or her needs. This system, referred to as content-based browsing, forms an abstraction to represent each shot of the sequence by using a representative frame, or an Rframe, and it includes management techniques to allow the user to easily navigate the Rframes. This methodology is superior to the current techniques of fast forward and rewind because rather than using every frame to view and judge the contents, only a few abstractions are used. Therefore, the need to retrieve the video from a storage system and to transmit every frame over the network in its entirety no longer exists, saving time, expenses, and bandwidth.

Model-based object recognition in dense-range images—a review

The goal in computer vision systems is to analyze data collected from the environment and derive an interpretation to complete a specified task. Vision system tasks may be divided into data acquisition, low-level processing, representation, model construction, and matching subtasks. This paper presents a comprehensive survey of model-based vision systems using dense-range images. A comprehensive survey of the recent publications in each subtask pertaining to dense-range image object recognition is presented.

Image processing on encoded video sequences

This paper presents a novel approach to processing encoded video sequences prior to complete decoding. Scene changes are easily detected using DCT coefficients in JPEG and MPEG encoded video sequences. In addition, by analyzing the DCT coefficients, regions of interest may be isolated prior to decompression, increasing the efficiency of any subsequent image processing steps, such as edge detection. The results are currently used in a video browser and are part of an ongoing research project in creating large video databases. The procedure is detailed with several examples presented and studied in depth.

Unsupervised classification of cell images using pyramid node linking

A segmentation technique which combines two properties in an iterative and hierarchical manner to correctly segment and classify the given cell images is described. The technique is applied to digital images taken from microscope slides of cultured rat liver cells, and the goal is to classify these cells into one of three possible classes. The cells of the first class (I) are morphologically normal and stain the darkest. Those of the second class (II) are slightly damaged, showing both nuclear and cytoplasmic swelling with resultant lessening of staining affinity. The cells of the third class (III) are markedly damaged, as demonstrated by the presence of cytoplasmic vacuolization, or are completely disintegrated. Cells of the first class are classified by taking advantage of the staining affinity; the original gray-level image is segmented into four gray levels. The darkest is then classified as type I. Type III cells are classified by using high busyness as a characteristic; the standard deviation of the original image is segmented into four busyness levels. The highest level is classified as type III cells. Assuming that only the three cell types are present in any given image, the remaining nonbackground unclassified pixels are determined to belong to type II cells.<<ETX>>

Hierarchical segmentation of 3-D range images

The authors present a novel approach for segmentation of dense three-dimensional range images. In this approach, four local properties, namely the 3-D coordinate, the surface normal, the Gaussian curvature, and the mean curvature of each data point, are combined in a hierarchical data structure to segment a given 3-D dense range map into surface patches. This algorithm is applicable to planar as well as curved surfaces; several examples of segmentation of such surfaces are presented.<<ETX>>

A Mechanically Derived Systolic Implementation of Pyramid Initialization

Pyramidal algorithms manipulate hierarchical representations of data and are used in many image processing applications, for example, image segmentation and border extraction. We present a systolic network which performs the first phase of pyramidal algorithms: initialization. The derivation of the systolic solution is governed by a mechanical method whose input is a known Pascal-like pyramidal algorithm. After a few manual program transformations that prepare the algorithm for the method, parallelism is infused mechanically. A processor layout is selected, and the channel connections follow immediately.

Convergence of fuzzy-pyramid algorithms

Pyramid linking is an important technique for segmenting images and has many applications in image processing and computer vision. The algorithm is closely related to the ISODATA clustering algorithm and shares some of its properties. This paper investigates this relationship and presents a proof of convergence for the pyramid linking algorithm. The convergence of the hard-pyramid linking algorithm has been shown in the past; however, there has been no proof of the convergence of fuzzy-pyramid linking algorithms. The proof of convergence is based on Zangwills theorem, which describes the convergence of an iterative algorithm in terms of a “descent function” of the algorithm. We show the existence of such a descent function of the pyramid algorithm and, further, show that all the conditions of Zangwills theorem are met; hence the algorithm converges.

Classification Of Complex Cell Images Using Pyramid Node Linking

There is a growing need to study and examine microscope slides in various fields of science. However, this task can be cumbersome and vulnerable to human error. This is especially true when large numbers of slides have to be examined. It is apparent that the existing computer and image processing technology should be utilized to speed up the process of cell examination. In our research, an existing method of image segmentation, called pyramid node linking, has been applied with a few modifications to cell segmentation. In pyramid node linking, a pyramid is constructed by successively reducing the resolution of the original image by factors of two to obtain the first and subsequent levels, until there are four pixels left on the top most level. In a bottom-to-top iterative process, the pixels from level to level are linked using information from the level above, the level below, and from the neighbor pixels on the same level. This results in several trees with roots at one of the upper levels and leaves on the original image. This process results in smooth simply-connected regions with well-defined boundaries on the bottom level. We have applied pyramid node linking to complex images consisting of clusters of rat liver cells grown in culture and damaged to different degrees by exposure to various chemicals. The algorithm has been applied to classify the rat liver cells in three categories: undamaged, slightly damaged, and disintegrated.