<span>Feature selection attracts researchers who deal with machine learning and data mining. It consists of selecting the variables that have the greatest impact on the dataset classification, and discarding the rest. This dimentionality reduction allows classifiers to be fast and more accurate. This paper traits the effect of feature selection on the accuracy of widely used classifiers in literature. These classifiers are compared with three real datasets which are pre-processed with feature selection methods. More than 9% amelioration in classification accuracy is observed, and k-means appears to be the most sensitive classifier to feature selection.</span>

No abstract

No abstract

No abstract

Text supply crucial suggestions for understanding video content, also the information that the text convey is much more concise than corresponding audio or video. The reason is that we need language knowledge to understand the text, and the knowledge itself does not need to be embedded in the text data. Text streams contain very rich semantic information. How to effectively extract information from text is an important component in video content analysis and semantics research. In this paper, a new morphology-based method for text detection in image and video is proposed. It consists of three major stages. In the first stage, the input color image is converted to gray-scale, a morphological binary map is generated by calculating the difference between the closing image and the opening image, and a binarization is performed. In the second stage, candidate regions are connected by using a morphological dilation and erosion operations. In the last stage, the extracted regions are verified based on characteristic of text regions to eliminate non text regions.