Median filtering: naive algorithm, histogram-based, binary tree, and more (2022)

The blog post discusses the concept of median filtering in image analysis. It explains how the median filter works by taking the median value of a pixel's neighborhood in the input image. The post delves into different filtering techniques such as percentile and rank filters, highlighting their applications in noisy image processing. Various algorithms for computing the median filter are explored, including a naive algorithm, a histogram-based algorithm, and a binary tree algorithm. The post compares the efficiency of these algorithms based on kernel size and complexity. Additionally, it introduces a constant-time algorithm proposed by Perreault and Hérbert for 8-bit images and square kernels. The post concludes with timing comparisons of different implementations, showcasing the computational efficiency of each method. Overall, the blog provides a comprehensive overview of median filtering techniques and their computational implications in image processing.