Point, Harris Corner, and SIFT Point Coordinate Encoding Algorithms
Jian–Jiun Ding, Szu‐Wei Fu, Pin-Xuan Lee · 2014
In this paper, we discuss a special data compression problem, i.e., how to compress the coordinates of points in an image. For the applications of template matching, object tracing, and object-oriented image processing, it is required to record the corners and the scale-invariant feature transform (SIFT) points of an object. The simplest way to encode these points is to record their coordinates directly. However, this method is inefficient. In this paper, we try a variety of methods to encode point coordinates. We find that the algorithm which adopts the techniques of shortest distance reordering, maximal axis distance, adaptive arithmetic coding, and dense-based context generation has the best performance to encode Harris' corners and SIFT points, the proposed algorithm is also useful for encoding other sparsely distributed two dimensional data.