• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.8 no.7
• /
• pp.1453-1463
• /
• 2004

The conventional image segmentation method using level set has been disadvantage since level set function in the gradient-based model evolves depending on the local profile of the edge. In this paper, a new model is introduced by hybridizing level set formulation and complementary smooth function in order to smooth the driving force. We consider an alternative way of getting the complementary function(CF) which is much easier to simulate and makes sense for most cases having no triple junctions. The rule of thumb is that CF must be computed such that the difference between their average and the original CF function should be able to introduce a reliable driving force for the evolution of the level set function. This proposed hybrid method tries to minimize drawbacks the conventional level set method.

• ETRI Journal
• /
• v.37 no.3
• /
• pp.541-550
• /
• 2015

In this paper, we propose a classification-based approach for hybridizing statistical machine translation and rulebased machine translation. Both the training dataset used in the learning of our proposed classifier and our feature extraction method affect the hybridization quality. To create one such training dataset, a previous approach used auto-evaluation metrics to determine from a set of component machine translation (MT) systems which gave the more accurate translation (by a comparative method). Once this had been determined, the most accurate translation was then labelled in such a way so as to indicate the MT system from which it came. In this previous approach, when the metric evaluation scores were low, there existed a high level of uncertainty as to which of the component MT systems was actually producing the better translation. To relax such uncertainty or error in classification, we propose an alternative approach to such labeling; that is, a cut-off method. In our experiments, using the aforementioned cut-off method in our proposed classifier, we managed to achieve a translation accuracy of 81.5% - a 5.0% improvement over existing methods.

• 한국균학회소식:학술대회논문집
• /
• 2006.10a
• /
• pp.32-44
• /
• 2006

To investigate non-aflatoxin-production of A. oryzae at the molecular level, an aflatoxin biosynthesis gene homolog cluster of RIB 40 was analyzed. Although most genes in the corresponding cluster exhibited from 97 to 99 % similarity to those of Aspergillus flavus, three genes shared 93 % similarity or less. In addition, although slight expression of aflR, positive transcriptional regulator gene, was detected in some A. oryzae strains having seven aflatoxin biosynthesis homologous genes, other genes related to aflatoxin production were not detected. RIB strains were mainly divided into group 1, having seven aflatoxin biosynthesis homologous genes (aflT, nor-i, aflR, norA, avnA, verB, and vbs), and group 2, having three homologous (avnA, verB, and vbs). Partial aflatoxin homologous gene cluster of RIB62 from group 2 was sequenced and compared with that of RIB40 from group 1. RIB62 showed a large deletion upstream of ver-1 with more than half of the aflatoxin homologous gene cluster missing including aflR, a positive transcriptional regulatory gene. Adjacent to the deletion of the aflatoxin homologous gene cluster, RIB62 has a unique sequence of about 8kb and a telomere. Southern analysis of A. oryzae RIB strains with four kinds of probe derived from the unique sequence of RIB62 showed that all group 2 strains have identical hybridizing signals. Polymerase chain reaction with specific primer set designed to amplify the junction between ver-1 and the unique sequence of RIB62 resulted in the same size of DNA fragment only from group 2 strains. Based on these results, we developed a useful genetic tool that distinguishes A. oryzae group 2 strains from the other groups' strains and propose that it might have differentiated from the ancestral strains due to chromosomal breakage.