Recognizing texts from camera images is a known hard problem because of the difficulties in text detection from the varied and complicated background. In this paper we propose a novel and efficient method to detect text region from images with complex background using Wavelet Transforms. The framework uses Wavelet Transformation of the original image in its grayscale form followed by Sub-band filtering. Then Region clustering technique is applied using centroids of the regions, further Bounding box is fitted to each region thus identifying the text regions. This method is much sophisticated and efficient than the previous methods as it doesn't stick to a particular font size of the text thus, making it generalized. The sample set used for experimental purpose consists of 50 images with varying backgrounds. Images with edge prominence are considered. Furthermore, our method can be easily customized for applications with different scopes.