ABSTRACT:
In the computer vision research area, the extraction and detection of text regions in images is a well known problem. In natural images, in order to extracting the text there are two basic approaches namely: texture based and edge based. A set of images of natural senses are evaluated and implemented through algorithms which are varying along with the dimensions of scale, lighting and orientation. In each process precision, accuracy and recall rates are analyzed for measuring the success and limitations of each and every approach. Based on the results recommendations are given for improving those areas where the results are very deprived. The images of text is assorted by means of camera from various distances i.e., scale variance is calculated for some of the images. In accordance with the lighting conditions images of texts are differed that means lighting variance is calculated for some images. For some images of text orientation variances are calculated and it is different from camera angles. Variances are calculated in each method i.e., texture as well as edge based methods. In this project, the main objective for calculating the variances are for which image which method is efficient as well as effective is identified.
No comments:
Post a Comment