Local directional coding-based scene change detection for frame rate up-conversion

Abstract

This paper proposes a new scene change detection method that uses local directional coding for frame rate up-conversion. The directional coding technique converts the luminance value of the edge direction response for given frames to bit codes and calculates the bit code difference between the previous and current frames. Existing scene change detection methods usually calculate the difference between the histogram shapes or the luminance values between the previous and current frames to detect scene changes. Thus, they erroneously detect scene changes in a scene with moving objects or camera motion. The existing automatic thresholding-based method uses iterative operations to calculate the optimal threshold and determine the scene change using the difference between the calculated threshold values, and it therefore has a high computational cost. To solve these problems, the proposed method uses the difference between the bit codes from the eight edge response values between the previous and current frames to detect the initial scene change regions. In addition, the proposed method uses a refinement process on the detected initial scene change regions to enhance the scene change detection accuracy. The experimental results showed that the proposed method enhanced average F1 score to 0.5235 (a 126.36 % improvement) as compared with the benchmark methods. The average computation time per pixel of the proposed method also reduced to 13.3668 μs (an 87.39 % reduction) compared with the benchmark methods.