An Improved Video Mosaic Block Detection Method on the Statistics of Pixel Difference across the Boundary of Macroblock and Suspected Mosaic Blocks

Abstract

This passage proposes a new method to detect mosaic not only using the Y (luminance) component in YUV color space of videos, but also using the U (chrominance) and V component. The mosaic effect is measured by the boundary pixel difference from the neighbor macroblock. Instead of detecting the existence and position of mosaic blocks as traditional methods do, this method focuses on the statistics of the number of suspected mosaic blocks so that the quality of the whole frame affected by mosaic can be assessed. Experimental results show that the new method has good performance on fallout ratio, omission factor and computational complexity.