The image data in most cases consists of two dimensinal data array, because it is the natural form for image input systems. While, some other data structues are expected to be more effective and convenient. The "image pyramid" and "image quadtree" are such structures. The image pyramid is the multi-layered data arrays with multi-resolutions. The image quadtree is an expression of black-and-white pictes using the four-branching tree. Some of the fundamental image processing algorithms have been developed on such data structures.
In pattern recognition, detection of partial ellipse is important, because a circlar object usually looks as ellipse unless seeing the object on its vertical center line, and it often becomes partial ellipse when it is hidden by some foreground object. However an efficient detection algorithm for a partial ellipse is scarcely reported except approximation by the least mean squares method. This paper described a newly developed detection algorithm for partial ellipse and some results of evaluation experiments.
The method of the cartesian-polar coordinate transform on digital images is described. To transform the digital image, it is necessary to give the method of sampring in the poler coordinate and the rule of mapping of pixels between both coordinates. The sampling method and the mapping rule are proposed so as to preserve the information included in cartesian pixels and to suppress the increase in the number of polar coordinate pixels. Experimental results show the effectiveness of the proposed method.
A high speed image processing micro-computer system is constructed using newly distributed LSI's ImPP (Image Pipelined Processor). This paper describes the configuration of the system and its application softwares-especially newly developed Maximum likely-hoodness classification program for ImPPs, including an efficent synchronizing method between parallel operated ImPPs. As a result, a remarkabie high speed classification is attained.
The encoding of the character information in English documents by the word-unit processing is described. This purpose is to improve the transmission efficiency in the mixed mode communication. This technique is composed of the following three processes. 1) A caharacter pattern extracted from the document image is recognized. 2) The rejected character is restored using the dictionaries based on the word formation (prefix, root, suffix) and the k-grams of the English language. 3) The recognized characters are encoded word by word by matching with the entries in the dictionaries. The computer simulation results for four A4 page documents show that about 4 times of compression ratio improvement is achieved compared with the MMR coding.
This paper describes the editing processing method to convert variable data from an information center into facsimile documents with geometric graphics such as graphs. The method assures high efficiency in processing by dividing a document into two parts : constant and variable. Constant part does not need editing process and is generated by the offline system. Variable part is generated by the variable data from an information center. Enhanced information providing services which use facsimile can be provided by the method.
An exact method determing optical flow from a sequence of images is presented. We adopt a pixel-based analysis calculating mutual-correlation functions between temporal change of brightness at a target pixel and that at neighboring ones. Local veloclty at the target pixel is deceided analytically from lag time of the correlation functions. The validity of the proposed method is confirmed by computer simulation of dynamic image.
A new velocimetry based on dynamic image processing is proposed. On the analogy of laser Doppler velocimetry, a dynamic scene is transformed to a sequential scene with sinusoidal field of gray-value. From temporal change of accumulated gray-value of each frame, we can evaluate a velocity information of moving particles using spectral analysis. The validity and usefulness of the velocimetry is confirmed using artificial images created by computer simulation.