Efficient implementation of codebook algorithm for background subtraction

Abstract

Background subtraction is a critical component of numerous image processing applications, particularly in video surveillance. Considerable effort has been devoted to enhancing the accuracy of codebook algorithms. However, these improvements have led to significant increases in computing power requirements, rendering them impractical for real-time applications. The performances of codebook algorithms are mainly due to its inherent complexity and the variable number of codewords dedicated to represents each pixel. Which requires frequent memory allocation and deallocation. To address these limitations, we propose optimizing the codebook model in order to reduce its size while preserving accuracy and enabling real-time processing.