<title>VLSI chip set for picture archiving and communication systems</title>

Abstract

Two Application Specific Integrated Circuits (ASIC) have been developed for hierarchical representation and compression of medical images for Picture Archiving & Communication System (PACS) environments. Hierarchical image representation enables progressive transmission of images and provides an economic means of rapidly surveying several images simultaneously. Furthermore, it facilitates the adaptation of high resolution images to displays with lower resolution. Compression enables efficient use of storage media and transmission channel capacities. Together the chip set implements a loss-less transform coding system.The first device implements the S-Transform algorithm (a 2x2 pixel case of the Hadamard Transform) which creates a hierarchical representation of an image. The second device implements the Lempel-Ziv compression algorithm, an adaptive codec that maps variable length strings of characters to fixed length code words. Used in tandem the chip set achieves image processing rates of up to 7.5million pixels per second at a system clock rate of 10MHz, for 13 bit pixel digitization. Both chips are microprocessor programmable. Because of the limited modes of the algorithm the S-Transform chip implementation focused on efficient utilization of silicon area and high processing rate. On the other hand because of its potential use outside of imaging applications, the Lempel-Ziv codec focused on versatility and speed. We describe the architecture and features of each chip and the characteristics of a system designed around these devices. I . INTRODUCTIONPicture Archiving and Communication Systems are providing a means for medical establishments to network their resources (equipment and physicians) and thereby increase the quality and efficiency of medical diagnostics. To achieve this requires that images be digitally acquired, stored, processed and transmitted. The typical x-ray image has a spatial resolution of 2K x 2K, and pixel digitization of 12 bits. To store such an image in its raw form requires at least 6.2M bytes of memory. To transmit the same image over a network with an effective bandwidth of lOOK bytes/sec takes approximately 1 minute. Considering the volume of images generated by hospitals, it is crucial that images be compressed for storage and transmission efficiency to lower the operating costs.Basically, there are two widely used methods by which compression can be achieved, predictive coding and transform coding1' 2• This paper relates to the latter method. Several transforms and coders have been employed for image processing applications. In this application the S-Transform3'4 forms the basis for hierarchical image decomposition. The Lempel-Ziv8'9 igorithm is the coder. Together, they provide efficient storage, transmission, and reconstruction of medical images.The main objective of this work is to realize the principal circuit elements as ASICs to facilitate their application in systems for image compression and decompression. Thus these chips substantially reduce the board area, power consumpt...