Vector Edge Detection in H.264 Implementation
Wei Zhao, Zuying Luo, Jeffrey Fan, Sheldon X.-D. Tan · 2008
Motion estimation (ME) is one of the bottlenecks in terms of the computational cost in a video encoder system. In this paper, we present a cost-effective method to calculate the "Vector Edge" of the current frame. We store all the vectors' information within a frame and put them in the Laplacian of Gaussian edge detection operator. It is similar to the image edge detection but we need to consider two vectors instead of one, which is different from the luminous edge detection.The experiments show that the proposed method is different from a luminance based graphic edge detection because the edge doesn't go off with high light or dark shader. The edge recognition is better in quality, based on the proposed two separate matrices along with their convolutions by the Laplacian of Gaussian operator.