National Library of Medicine, Office of Acquisitions Management, Building 38A, Room B1N17, 8600 Rockville Pike, Bethesda, MD 20894

A -- VISIBLE HUMAN IMAGES COMPRESSION SOL NLM 97-062/VMS DUE 060997 POC Valerie M. Syed, Contract Specialist (301) 496-6546 The National Library of Medicine intends to negotiate on a sole source basis with Texas Tech University, Office of Research Services, 203 Holden Hall, Lubbock, Texas 79409 for professional services to support ongoing research in the compression of the Visible Human images over a 9-month period beginning June 16, 1997. NLM's Visible Human (VH) cryosectioned 24 bit/pixel color images are highly data intensive, and challenge both transmission and storage. Our objective is to develop a system that can provide the best means of storing this image set without losing original data, and providing access to it via the Internet at a lossiness controlled by the requester (user). This strategy, therefore, is based on a user-centered philosophy, not the conventional approach of compressing images by multiple techniques, and allowing experts to examine the expanded images for subjective quality. This task requires a statistical analysis of a representative sample of the color VH images, based on which the optimum wavelet transform, optimum number of taps, and the levels of wavelet-based multiresolution decomposition levels, to decompose the VH image set are selected. Also, the quantization method (scalar, vector) and number of quantization levels that are suitable in this application shall be selected. Candidates for the quantization method shall include the scalar technique known as EZW and the adaptive vector quantization techniques (AFLC and IAFC). The AFLC and IAFC techniques were developed by Prof. Sunanda Mitra at Texas Tech University and are unique to this university. Criteria for optimality shall include objective quality measures that may include any or all of the following: L1 error, L2 error, root mean square error (RMSE), peak signal/noise ration (PSNR) and maximum error. The selected wavelet transform/quantization technique shall be expressed in the form of algorithms suitable for coding. The task also includes the development of executable software modules in C and/or C++ languages to implement the algorithms for both the wavelet transformation of the images and the subsequent quantization. The modules shall be applied to a sample of the VH images as selected by the Government, the numbers of levels shall be varied, and compression ratio shall be measured for different combination of decomposition levels and quantization levels. Also, the performance of the quantization technique shall be determined by the rate distortion function generated by the quantization algorithm and by weighting the wavelet coefficients with the human visual system model. Finally, the task shall include the development of a functional design for the client software required for the user to select quality level desired, access the data store, and retrieve images either singly or in a set, both access and image transmission to be done via the Internet. This notice of intent is not a request for competitive proposals. Organizations may identify in writing their interest and capability in response to this requirement or submit a proposal.However all responses received no later than the closing date of this announcement will be considered by the government. A determination by the government not to compete this proposed acquisition based upon responses to this notice is solely within the discretion of the government. Information received will normally be considered solely for the purpose of determining whether to conduct a competitive procurement. No solicitation document exists, therefore an analysis will be made on the basis of information received in response to this notice. Responses are required within fifteen (15) calendar days from the publication date of this notice to The National Library of Medicine, Office of Acquisitions Management, 8600 Rockville Pike, Bldg. 38A, Room B1N17, Bethesda, MD 20894, Attention: Valerie M. Syed. (0142)

Loren Data Corp. http://www.ld.com (SYN# 0004 19970527\A-0004.SOL)