针对尺度不变特征变换 (Scale-invariant feature transform, SIFT) 算法在关键点特征描述向量阶段计算复杂并且维数较高的现象,提出了一 种基于压缩感知理论的SIFT算法。通过压缩感知理论的稀疏特征表示方法,对SIFT关键点特征 向量进行提取,将高维梯度导数向量降到低维稀疏特征向量,降低了关键点描述向量维度。采 用欧式距离作为关键点的相似性度量, Best-Bin-First(BBF) 数据结构避免穷举,使数据的运算量大为减少。实验结果表明,新算法对存 在仿射变换的医学图像配准性能优于传统SIFT算法,与当前改进型的SIFT算法相比,本文算法的实时性明显增强。

Scale-ivariant feature transform (SIFT) has defects in computational complexity of its key point descriptor computing stage and in the high dimensionality of the key point feature vectors. To speed up the computation, a SIFT based on compressive sensing algorithm was proposed. By the sparse feature representation methods of compressive sensing theory, the feature vector of SIFT was extracted and the high-dimensional gradient derivative was decreased to low-dimensional sparse feature vector. Accordingly, Euclidean distance was introduced to compute the similarity and dissimilarity between feature vectors used for image registration and Best-Bin-First (BBF) data structure was used to avoid exhaustion. The experimental results show that the proposed algorithm has better performance than the standard SIFT algorithm while registering the affine transformation medical images. Comparing with the current modified SIFT algorithms, the real-time performance of the proposed algorithm is improved obviously.