A related problem is the storage of microscopic images. The National Institute of Health (NIH) in America issued a call for proposals on this problem in April 2009. Recent advances in technology have enabled researchers to digitize pathological whole-slides at high magnifications (e.g. at 40). Resulting colour microscopic images are large in size (e.g. 100K by 100K). Therefore, finding efficient ways of compressing such images is necessary for efficient transmission, storage and evaluation.
Currently, standard image compression methods such as JPEG or JPEG2000 are used. However they do not take full advantage of the specific nature of microscopic images. For example, IHC and H&E images contain textures that are “impulsive” in nature, as shown in Figure-1 and their colour content depends on the staining process. Therefore, compression methods suitably tailored for these images should be developed. In this WP, we will address both the colour and the texture aspects of the image compression problem.

Recently, OSU and BILKENT developed a histogram-based colour transformation method for microscopic images, that takes the image specific colour-content into account and can be integrated with standard image compression methods like JPEG. The conducted tests showed that, when applied to microscopic images, the proposed method yields a mean compression ratio gain of about 20% for a given PSNR value, compared to standard JPEG. Further compression with the JPEG and JPEG-2000 will be carried out in this WP.
Image compression schemes for impulsively textured microscopic images will be developed. We developed adaptive wavelet transform methods for isolated points in 3D data [26,27]. It should be pointed out that all of the partners have experience in wavelet analysis. We will develop an Adaptive Wavelet Transform (AWT) [28,29] method for impulsively textured microscopic images. The method will take advantage of the correlation between isolated impulsive pixels as in [26,27] and this will lead to more efficient data compression results compared to the ordinary wavelet transforms. We plan to encode the AWT coefficients using an embedded zero tree. In this respect, Dr. Rajpoot’s expertise in adaptive wavelets and zero-tree based methods for compression of images [30,31,32] will be utilized for developing efficient histology image storage algorithms.