A design of an associative memory array processor for ultrasonograph image acquisition and processing

This paper describes a design of an associative memory array processor that can be used in the acquisition and processing of ultrasonograph images. The major concept is to design a parallel architecture that reduces task's execution time by analyzing multiple parts of the image concurrently. The architecture constitutes a distinctive type of single-instruction stream, multiple-data stream machine that is built around content-addressable associative memory slabs, that allow parallel access of multiple memory words. The basic building block of this architecture is a one-pixel processing element, which can perform the standard load (data acquisition) function and also contains some special comparison logic to enable its content to be compared with an external data. Several image processing operations are implemented in parallel, among them: component labeling, size filtering, pattern centralization, and pattern recognition. The proposed novel architecture can label specific regions into the image and isolate them intelligently. It is also capable of storing templates that may be considered as references for similar cases. The system is able to perform learning process and extract features from several input patterns and store the reference pattern in a slice. Moreover, the system is capable of comparing an input image with a pre-stored template during recognition process. The proposed architecture is of interest because it speeds up the recognition process and helps radiology specialists to write their reports confidently.