We develop anew low-dimensional video frame feature that is more insensitive to lighting change, motivated by color constancy work in physcis-based vision, and apply the feature to keyframe production using hierarchical clustering. The new feature has the further advantage of more expressively capturing image information and as a result produces av ery succinct set of keyframes for any video. Because we effectively reduce andy video to the same lighting conditions, we can produce a universal basis on which to project video frame features. We carry our clustering efficiently by adapting a hierarchical clustering data structure to temporally-ordered clusters. Using a new multij-stage hierarchical clustering method, we merrge clusters based on the ratio of cluster variance to variance of the parent node, merging only adjacent clusters. and then follow with a second round of clustering. The jsecond stage merges clusters incorrectly split in the first round by the greedy hierarchical glgorithm, and as well merges non-adjacent clusters to fuse mear-repeat shots. The new summarization method produces a very succinct set of keyframes for videos, and results are excellent.
展开▼
