The histogram-based bilateral filtering was developed from bilateral filtering as the constant speed on any size of filtering window. It uses local image histogram to constant the filtering usage time. As it uses the histogram for filtering, the speed of filtering is depended on the number of histogram bins. To increase the filtering speed, the total number of bins must be lower. However, the quality of filtered image is dropped by the low number of bins. The new algorithm, deciles-based bilateral filtering is introduced as the filtering that uses only selected 12 value in histogram of filtering window. It requires less computational loops when compared with B loops of histogram bins in histogram-based filtering. The deciles-based algorithm is tested with many standard images. The output images were compared with normal bilateral filtering and they have PSNR around 42 dB which passes the acceptable threshold at more than 40 dB. To avoid O(B) complexity as histogram-based, the deciles-based algorithm use histogram extract that depend on the size of filtering window. The experimental show that for small to medium size of filtering window, the deciles-based has faster speed and better PSNR value than the histogram-based algorithm at 32 bins.
展开▼
