July 2011
Intermediate to advanced
648 pages
13h 2m
English
Optical flow is a fundamental concept in computer vision. Optical flow describes the motion that occurs between images and has application in image stabilization, frame rate upsampling, and motion or gesture analysis. In this chapter, we will discuss an implementation of pyramidal Lucas-Kanade (LK) optical flow.1 Sub-pixel accuracy is essential to the implementation and is maintained by a number of resampling operations requiring linear interpolation. As we will see, the texture-filtering hardware on the GPU can be used to perform linear interpolation of data, which can provide significant speedups. Additionally, we will discuss how local memory and early kernel exit techniques provide acceleration.
Read now
Unlock full access