WalshHadamard Transform Kernel- Based Feature Vector for Shot Boundary Detection
- Lakshmi Priya, G.G, S Domnic
- IEEE Transactions on Image Processing,
Shot boundary detection (SBD) is the first step towards video indexing and content based video management. Due to availability of low cost storage media devices, and broadband data connection, digital videos are becoming widely used. However, the increasing availability of digital video has not been accompanied by an increase in its ease of accessibility. If we want to see a clip of interest, we have to sequentially browse through the video. This is an extremely time consuming and tedious process. So, accurate shot boundary detection plays vital role to organize video contents into meaningful parts for video scene analysis. A shot is defined as an unbroken sequence of frames taken from camera. Shot transitions can be either abrupt (cut) or gradual (fades, dissolves, wipes). In this paper, a new shot boundary detection (SBD) method is proposed using Structural SIMilarity (SSIM) Index. The abrupt cuts are identified using SSIM and gradual transitions(fades) are identified using standard deviation plot of the frames in the video. The proposed method only needs mean, standard deviation and co-variance of the frames as basic input parameters for detecting cuts and gradual transitions. The performance of the proposed method is comparable with that of the existing global and local histogram method for SBD.