Abstract:

For the linear edges,the method using the traditional Sobel operator for edge detection and direction discrimination can not accurately detect the direction.An algorithm for linear edge detection using Hessian matrix with half pixel precision is proposed.The pixels in the lost block of the image are classified,for different edges, we use different edge detection operator to detect the directions of the edges.The corrupted image is concealed by directional extrapolation according to the detected edge direction.The experimental results show that Hessian matrix with half pixel precision for linear edges detection can get more precise direction,the average PSNR(Peak Signul Noise Rate) of the reconstructed image is 0.2~0.4 dB higher than that of other existing techniques.

Key words: half pixel, Hessian matrix, edge detection, directional extrapolation, spatial error concealment