Block-based connected components labeling is by far the fastest algorithm to label the connected components in 2D binary images, especially when the image size is quite large. This algorithm produces a decision tree that contains 211 leaf nodes with 14 levels for the depth of a tree and an average depth of 1.5923. This article attempts to provide a faster method for connected components labeling. We propose two new scan masks for connected components labeling, namely, the pixel-based scan mask and the block-based scan mask. In the final stage, the block-based scan mask is transformed to a near-optimal decision tree. We conducted comparative experiments using different sources of images for examining the performance of the proposed method against the existing methods. We also performed an average tree depth analysis and tree balance analysis to consolidate the performance improvement over the existing methods. Most significantly, the proposed method produces a decision tree containing 86 leaf nodes with 12 levels for the depth of a tree and an average depth of 1.4593, resulting in faster execution time, especially when the foreground density is equal to or greater than the background density of the images.

Quick Response (QR) codes seem to appear everywhere these days. We can see them on posters, magazine ads, websites, product packaging and so on. Using the QR codes is one of the most intriguing ways of digitally connecting consumers to the internet via mobile phones since the mobile phones have become a basic necessity thing of everyone. In this paper, we present a methodology for creating QR codes by which the users enter text into a web browser and get the QR code generated. Drupal module was used in conjunction with the popular libqrencode C library to develop user interface on the web browser and encode data in a QR Code symbol. The experiment was conducted using single and multiple lines of text in both English and Thai languages. The result shows that all QR encoding outputs were successfully and correctly generated.

In this paper, we investigate predicting the Stock Exchange of Thailand Index movement. Currently, there are two stock markets in Thailand; the Stock Exchange of Thailand (SET) and the Market for Alternative Investment (MAI). This paper focuses on the movement of the Stock Exchange of Thailand Index (SET Index). The back propagation neural network (BPNN) technology was employed in forecasting the SET index. An experiment was conducted by using data of 124 trading days from 2 July 2004 to 30 December 2004. The data were divided into two groups: 53 days for BPNN training and 71 days for testing. The experimental results show that the BPNN successfully predicts the SET Index with less than 2% error. The BPNN also achieves a lower prediction error when compared with the Adaptive Evolution Strategy, but a higher prediction error when compared with the (1+1) Evolution Strategy.

No abstract

Named Entity (NE) extraction for Thai language is a difficult and time consuming task because sentences in Thai language are composed of a series of words formed by a stream of characters. Moreover, there are no delimiters (blank space) to show word boundaries. Currently, most named entity extraction methods for Thai language are associated with word segmentation and Part of Speech (POS) tagging processes. The accuracy of named entity extraction is mostly affected the efficiency of those processes. At present, it is still lack of suitable methods for identifying the boundary of word for Thai sentence. Therefore this paper proposes the method to extract Thai personal named entity without using word segmentation or POS tagging. The proposed method is composed of 3 steps. Firstly, pre-processing, this process is used to remove non alphabet such as parenthesizes and numerical. Then, personal named entity is extracted by using contextual environment, front and rear, of personal name. Finally, post-processing, a simple rule base is employed to identify personal names. The training corpus of 900 political news articles and the test corpus of 100 political news, 100 financial news and 100 sport news articles were used in the experiments. The results showed that the F-measures in political and financial domain are 91.442% and 91.720% respectively which are nearly the same as in [5]. However, the proposed scheme used neither word segmentation nor POS tagging process that can significantly reduce the effort and speed up the process in building the training corpus.