In computer vision, the corners of an object play an important role in shape representation and analysis. In this paper, we carried out a dozen popular and most cited corner detection algorithms and studied their performances based on several performance measures proposed by us as well as in the literature over a varied range of images and categorized the corner detectors based on different image types and rank their performances with suitable threshold ranges corresponding to different image types. After obtaining a suitable corner detector for a given type of images, the paper describes a new approach to corner detection in a digital image based on the assumption that corners are those image points with high information content, and hence corners in an image exist in those regions having considerably high-intensity variation. Consequently, a complex corner response function is computed only within those regions with considerable high-intensity variation instead of entire image, reducing the computational cost of the whole procedure. Experiments conducted with the help of a few images showed the efficiency of the technique, both in terms of execution time and false-positive corners.