Keywords : Correlation


Remove Unimportant Features from True Colored Images Using the Segmentation Technique

Shahad A. Hasso

AL-Rafidain Journal of Computer Sciences and Mathematics, 2010, Volume 7, Issue 3, Pages 121-134
DOI: 10.33899/csmj.2010.163932

In this work a new approach was built to apply k-means algorithm on true colored images (24bit images) which are usually treated by researchers as three image (RGB) that are classified to 15 class maximum only. We find the true image as 24 bit and classify it to more than 50 classes. As we know k-means algorithm classify images to many independent classes or features and we could increase the class number therefore we could remove the classes or features that have minimum number of pixels which are considered unimportant features and reconstruct the images.
Correlation factor and Signal to Noise Ratio were used to measure the work and the results seems that by increasing the image resolution the effect of removing minimum features is decreased.
The CSharp (Visual Studio 2008) programming language was used to build the algorithms which are able to allocate huge matrices in high execution time.
 

The Effects of Correlated Data and Correction Procedures for F-Test in Unbalanced Two Way Model

Ivan S. Kababchi

AL-Rafidain Journal of Computer Sciences and Mathematics, 2010, Volume 7, Issue 2, Pages 13-28
DOI: 10.33899/csmj.2010.163875

          Independence of observations is one of the standard assumption in analysis of Variance (ANOVA) table. Where the error terms in the model are independent, identically distributed normal variables with null means and homogeneous variances. In this paper investigate the effect of dependence of observations in ANOVA for unbalanced 2-way nested fixed  model and developing a method for adjusting it. When the error terms are correlated and focus on the effects of departures from independence assumptions on hypothesis testing by determining the expect mean squares for errors as well as treatments for this model and correcting the F statistics for testing the factor effect. The model considered is one in which all measurements have same variance, and the covariance matrix enjoy a structure defined as follows: every pair of measurements comes from:
i) The same experimental observation and the same experimental unit;
ii) Different experimental observation, but in the same experimental unit;
ii) Different experimental unit;
has covariance  and   respectively.