A Study of Image Enhancement of Cerebral DSA
脑部数字减影图象增强研究

A serious disadvantage of Digital subtraction angiography(DSA), inherent to the subtraction operation, is its sensitivity to patient motion. The resulting artifacts frequently reduce the diagnostic value of the images. This paper presents an algorithm based on registration for the image enhancement of DSA, it employs block matching method for the registration of mask image and live image. Because of the high computational cost, it is usually not admissible to determine the displacement explicitly for every pixel. This has led to the concept of control-point based registration. Most control-point based registration algorithms published so far employ regular grids. However, control point selection should be image-feature based. Our experiments have demonstrated the advantage of algorithm with edge based control point selection over algorithms based on regular grids. Other aspects in our algorithm include similarity measure, interpolation and search strategy: energy of the histogram of difference is selected as the similarity measure, fast spline interpolation is used in order to implement the algorithm fast and get better result, and Powell optimization strategy is used as the search strategy by carefully choosing block size. The experiments demonstrate that the subtraction images after registration are of high quality and much more diagnostic value, the algorithm is also fast to be implemented.