Image data compression using a new floating-point digital signal processor

A new dual-ported, floating-point, digital signal processor has been evaluated for compressing 512 and 1,024 digital radiographic images using a full-frame, two-dimensional, discrete cosine transform (2D-DCT). The floating point digital signal processor operates at 49.5 million floating point instructions per second (MFLOPS). The level of compression can be changed by varying four parameters in the lossy compression algorithm. Throughput times were measured for both 2D-DCT compression and decompression. For a 1,024×1,024×10-bit image with a compression ratio of 316∶1, the throughput was 75.73 seconds (compression plus decompression throughput). For a digital fluorography 1,024×1,024×8-bit image and a compression ratio of 26∶1, the total throughput time was 63.23 seconds. For a computed tomography image of 512×512×12 bits and a compression ratio of 10∶1 the throughput time was 19.65 seconds.