The extraction of text in an image is a classical problem in the computer vision. Extraction involves detection, localization, tracking, extraction, enhancement and recognition of the text from the given image. However variation of text due to difference in size, style, orientation, alignment, low image contrast and complex background make the problem of automatic text extraction extremely challenging. Text extraction requires binarization which leads to loss of significant information contained in gray scale images. The images may contain noise and have complex structure which makes the extraction more difficult. This paper proposes an algorithm which is insensitive to noise, skew and text orientation. It is free from artifacts that are usually introduced by thresholding using morphological operators. Examples are presented to illustrate the performance of proposed method. The text extraction system has been attempted over a corpus of three kinds of images and promising precision has been obtained.