Video partitioning by an illumination invariant metric based on edge map

Automatic video partitioning is the first step for content-based parsing and indexing of video data. Many methods have been introduced to address this problem, e.g. pixel-by-pixel comparisons and histogram comparisons. Histogram is robust to object motion, therefore it is used widely, but it is sensitive to illumination variation inherent in the video production process especially in TV news reports. So false alarms are inevitable. The color ratio histogram, which is robust to illumination changes, is proposed as a frame content measure to solve the problem. However it is computationally expensive. A new illumination invariant metric to measure the content of image is proposed to segment video. The metric is based on the edge map of the images. It is robust to object motion, camera movement as well as illumination variation but it is sensitive to both camera break and gradual transition. No false alarm is made due to illumination variation. The metric is more economic regarding the computing cost since the ratio histogram and gray/color histogram are vectors while area is simply a scalar quantity. The effectiveness of our method has been validated by experiments on some real-world video sequences