Path: utzoo!mnetor!tmsoft!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!uakari.primate.wisc.edu!aplcen!boingo.med.jhu.edu!cortex!jamesr
From: jamesr@cortex.med.jhu.edu (Jim Rosenberg)
Newsgroups: comp.databases
Subject: Re: database containing gray-scale graphic images
Summary: Blob-Handling RDBMS's for Unix (Ultrix) systems ? Speed of access ?
Keywords: gray-scale graphic images blobs
Message-ID: <1094@cortex.med.jhu.edu>
Date: 20 Dec 90 15:53:48 GMT
References: <1990Dec19.181537.10103@usenet.ins.cwru.edu> <117@intrbas.UUCP>
Organization: Psych. Neuro-Imaging, Johns Hopkins
Lines: 52

In article <117@intrbas.UUCP>, kenn@intrbas.uucp (Kenneth G. Goutal) writes:
> InterBase is a relational dbms that runs on SCO Unix and SCO Xenix
> and can handle graphics perfectly well, in fields of a type called a 'blob'.

Will Interbase run on a DEC workstation network (3100's running off of
a 5400 server) running Ultrix ?  Are there other RDBMS's that handle
gray-scale images that will run on such a network ?  How well do
Interbase and any other RDBMS's with "blobs" handle multi-vendor networks,
and networks that combine PC's and Unix workstations ?

> Once a record has been selected, the field can be read with getb()
> can also read large chunks of a blob called 'segments'... for efficiency
> blob_dump()... dumps the field into a real Unix file,
> A parallel set of functions are available for writing the field

How is the speed on retrieving and storing images in these RDBMS's ?  
Are there ANY limits on the size of the "blob."  Is there a clear 
relationship between performance and the size of the blob (aside from 
system throughput constraints) ?  Does the inclusion of "blobs" affect 
the performance of the RDBMS in handling queries / updates of "traditional"
domains such as "name" or "age" ?  Can calls for "blobs" be embedded in "C"
programs ? How do these RDBMS's compare with "non-blob" RDBMS's in general 
(query languages, 4GL's, form makers and report writers) ?

We are a medical imaging research lab with approx. 2500 image series on 
approx. 1000 subjects.  An "image series" consist of between 10 and 60
individual images, all in a file between 3Mb and 10Mb.  I am currrently 
designing a RDB for all the information we have on these subjects 
(test scores, demographics, etc.).  I am interested in the possibility 
of using a RDBMS to manage access to the medical images as well; this 
would require embedding queries for the images into "C" and "X Windows" 
programs we use for analysis.  Speed of access is the critical factor in
determining the feasibility of using an RDBMS:  our users need those images
in (very few) seconds.

> We also have a notion called 'blob filters'... functions you... define...
> whenever it's appropriate to convert from one kind of a blob to another...
> perhaps converting some internal representation of gray scale to
> different formats for different... hardware and/or software.

This sounds great !  Is there this sort of support in other "blob handling"
RDBMS's ?

Thanks in advance for all the information.  Response to this news group
or by email (jamesr@cortex.med.jhu.edu) is fine.  If there is a lot of 
interest and I recieve a number of email responses, I will summarize to
the net.

James Rosenberg (jamesr@cortex.med.jhu.edu)
Systems Manager
Div. Psych. Neuro-Imaging
Johns Hopkins Medical Institutes