Abstract

Image processing tools and libraries have been written to analyze images for the properties of objects within.  Many image processing tools require human intervention to interpret them, and have no preview capability.  Progress made in a course for pattern recognition has revealed a basic methodology for simple classification of images.  

One application of these techniques are FITS images as recorded in the Data Release 1 of the Sloan Digital Sky Survey housed at Texas Tech University.  A set of Web Objects applications and services provide means of mapping images to their sky coordinates.   

The DCG FITS Services provide a bridge between the FITS images to Core Image.  The services further provide meta-data information for analysis and use image reproduction.  Next, these non-kernel and kernel methods provide basic image segmentation and de-noising.  Lastly, a Cocoa interface is supplied to simplify to road from desired analysis to discovery.  All of this just to examine basic principals of pattern recognition on images and the sky.

Construction of these frameworks use both non Core Image kernels and some preliminary work Core Image kernels for basic image segmentation.

Introduction
    
    
An image query tool was motivated by the loss of the original Sloan Digital Sky Survey Data Release 1 (DR1) repository in either of its original forms at Fermi National Accelerator Lab (FNAL), which was replaced by Data Release 4 and 5.  The mirror at Texas Tech is the only copy of DR1 left in its original form, and the means of its retrieval is what this paper is about.  
DR1 had some properties that were intended to be refined into a content management system for Flexible Image Transport System (FITS) images and spectrographs.  A demonstration of this tool has been developed and is located at http://venus.cs.ttu.edu/sdss (under tool). 

The original IQS tool used the same database as the tool shown in this report but had far more complex queries.  IQS produces URL(s) for the pre-computed JPEG image and original FITS images.  IQS accomplishes this task via Enterprise Object (EO) Model\cite{mendis}, which may be used for both web applications and generalized web services.

Overview

The first attempt of generating a content management system (CMS) for  FITS was at large based on the FITS format itself.  FITS consists of blocks called header data units (HDU), composed of a dictionary-like header and data unit that may be an image or table.  Each HDU is allocated in 2880 byte blocks.   Each header entry occupies 80 bytes within a header block.  The header is not a true dictionary as it allows for an entry to occur more than once.  

One of the complementary models is an Objective-C wrapper for the CFITSIO library supplied via HEARSAC \cite{heasarcfv}, a research group at NASA.  The first attempt for using this library was incomplete.  It left out significant frequency domain and scaling factors necessary for proper color arrangement of the image.   This fact was accidently discovered on a histogram study of sample FITS images in the FV application (a UNIX application written by HEARSAC product) and the results from a test application of the wrapper called FITS-OBJC.  In actuality, FITS supplies a rich set of dictionary values for image processing filters such as exposure.     Subsequent studies of FITS-OBJC revealed that the exposure value was the key value altered, and that Core Image's\cite{apple-core-image} exposure filter could be used to properly restore the image. 

One simpler and fundamental means of using the FITS files is obtaining them in an orderly fashion.    This is where the DR1, a demonstration of IQS, comes in.  IQS acquires references to the files  in two ways.   First, it provides a mapping of astronomical coordinates to the coordinates of observations recorded by SDSS DR1.  Second, IQS maps these results to URL references of the data files in the repository. 

IQS also demonstrates some of the fundamentally straight forward mechanisms provided in Apple's Web-Objects framework.  Nearly synonymous with Web-Objects are the Enterprise Objects framework (EO framework) which provide the necessary database access and intermediary actions for this clean model.  It is hoped that this model may aid a complementary model for interpreting the FITS structure itself, and any processing inquiry derived.  

Enterprise Objects contribute to reassembling data acquired in the DR1 collection.  Originally, DR1 was assembled with meta-data from the data samples collected, namely images and spectrographs.  Meta-data acquired through image processing was used in selecting objects for spectrographic observation.   Those attributes were defined in processes of Near Field (NF) Calibration, astrometric calibration, correlated image references, and other calibration data\cite{fpAtlas-sdss-dr4},\cite{asTrans-sdss-dr4},\cite{raddick-cmd}.   The calibration data maps the essential sample reference metrics (run, rerun, field, and camera column) and astrometric  reference for each object within.   The astrometric reference includes its standard coordinate data.   The standard coordinate to reference metric is crucial for identifying both the objects in the SDSS data repository, and determining which reference metric has the object desired.  This data is contained in the DR1 database as a table called Spectral Object (specobj)\cite{dr1-sdss}. 

The Spectral Object table is large since it includes most of the meta-data of both the spectral analysis and image processing used to select the object to be observed.  The EO model constructed to generate a simple demo was named Compact Spectral Object (compactSpecObj.eomodel in implementation).  It is defined on the database in terms of right ascension, declination, run, rerun, camera column and field.  Right ascension and declination are what any astronomy student can use to identify any point in the celestial sphere.  

The first version of an IQS web application is located on the web at http://venus.cs.ttu.edu/cgi-bin/WebObjects/iqTestCooridinateOriented.woa .   It uses two inputs (right ascension and declination).  The output is placed on to a template for each non-duplicate result.  The template assigns links for the pre-computed JPEG images, the $\frac{1}{15}$ scale JPEG image is assigned to the preview image.   The original FITS image link is assigned to each of the five color links below the preview image.   

In the event that the user choses a region of the sky that has not been observed by DR1, the results will be empty thus no set of preview templates will show up.  The entry boxes for RA and DEC are still provided.  

Conclusions:

IQS in its web application form demonstrates the use of Enterprise Objects and how it greatly simplifies the development of scientific content management systems. What is extraordinary about IQS is that its web services form is derived entirely from the same EO model as the application is built from. 

IQS does have its limitations, but as a building block it has profound ramifications.  Its supplied URL's can be used in its web application form to acquire any image supplied in the DR1 repository at Texas Tech.  Its web services form can supply the URL(s) for any of the data contained in DR1 and thus allow a client to have access to this data.