DNA microarray images, which are the final results of a microarray experiment, are usually massive in size. We have designed and implemented an algorithm for lossy compression of microarray images. First, spots are automatically extracted into spot sub-images. Then, within each subimage, a circle is matched to each spot. This is done via spatial optimization techniques. After that, our circle to square transform, C2S, is applied to the sub-image and a square image representing its spot part is achieved. Information regarding the coordinates of the C2S transform is also included in the final compressed file. The background parts which contain no biological information are neglected. The square shaped images are then put together and coded by means of DCT transform, quantization, and entropy coding.