The cDNA microarray image technology is a powerful tool for monitoring the expressions of thousands of genes simultaneously. An experiment is comprised of hundreds of images, each image easily over 30MB. Since image processing and statistical analysis tools are still under development, the images are always kept. Current focus on the development of standards makes efficient data transmission an important problem. Though the cost of disk space for storage is decreasing, efficient transmission requires compression. We have developed a partially progressive compression scheme for microarray images, allowing for fast decoding and reprocessing of image subsets. The scheme also permits locally varying image distortion or loss. The degree of loss can be chosen on-line, or be based on local parameters such as the spot intensities, or signal-to-noise ratios. The minimum decodable bitrate depends on the initial choice of parameters. We find that a bitrate of 4.1 bpp (cmp 32 bpp uncompressed) is sufficient for most tasks, such as image segmentation, and gene expression level extraction with a variety of existing methods. The lossless bitrate is about 17.5 bpp, comparable to the state-of-the-art lossless schemes, yet with the added flexibility of a progressive scheme. Our scheme has been tested on microarray images from different labs, and on images of varying quality.