At the April 2000 Windows Hardware Engineering Conference (WinHEC; New Orleans, LA), Pat Gelsinger, vice president of Intel desktop products group, announced the release of the Universal Serial Bus (USB) 2.0 specification. Although USB 2.0 is designed to extend the speed of the peripheral-to-PC connection from 12 Mbits/s on USB 1.1 to up to 480 Mbits/s, designers of image scanners still must deal with the challenges associated with moving very large color images over such interfaces.
EMBEDDING COMPRESSION HARDWARE
To deal with these problems, designers are looking to embed image-compression hardware in image scanners before transmission over serial interfaces. In designs such as medical-image scanners, such lossless image compression can reduce image-transfer time by a factor of two or more. In desktop color-image scanners, where lossless compression is not a requirement, this level of compression can be further increased.
Realizing this, Advanced Hardware Architectures (AHA; Pullman, WA) has introduced its AHA3710, a single-chip lossless/lossy compression coprocessor integrated circuit (IC) that supports both lossless and lossy compression for 24-bit color image data or 8-bit gray-scale data. Developed for applications requiring a compression-only device that compresses color data at a maximum rate of 6 Mbytes/s, the IC is capable of image compression between 2:1and 4:1.
Designed as an embedded processor, the AHA3710 co-processor IC can be used to compress: (left) 8-bit gray-scale scanned text images, (middle) 8-bit continuous-tone composite gray-scale images, and (right) 24-bit color (RGB) continuous-tone images. The compression performance of each algorithm is listed.
In lossless compression mode, a file can be compressed and then decompressed to recover the original file. Typical lossless compression ratios are 2:1 for most images. In lossy compression mode, some resolution is lost, depending on which lossy compression algorithm is selected. As the resolution is reduced, the resulting compression ratio is increased.
The lossless color-adaptive data-compression (CADC) algorithm is based on differential pulse-code modulation and Huffman coding. For red-green-blue (RGB) formatted color images, the data are deinterleaved into separate color planes and each plane is processed individually.
After compression and transmission, PC-resident images are decompressed using CADC software that is supplied free of charge or licensed from AHA. This software decompresses any data processed through the AHA3710 IC, including both lossless and lossy compressed files with any valid quantization factors. Because the decompression software automatically determines the compression parameters, no values need to be passed on to the decompression software. Optimized for Intel CPUs, typical decompression data rates exceed 2.0 Mbytes/s (depending on the microprocessor speed and the compression ratio).