Streaming-Mode MB-Based Integral Image Techniques for Fast Multi-view Video Illumination Compensation

Abstract

Multi-view video systems are often faced with brightness variations across the multi-perspective captured images. To tackle this problem and maintain the coding performance, block-based illumination compensation (BBIC) methods are recently proposed, where a first order affine BBIC model, consisting of a multiplicative factor and an additive offset, is often adopted. However, so far little attention has been paid to the fast algorithms that can reduce the computational overhead of BBIC. Therefore, we propose a fast image local statistics computation scheme using the technique of integral images, which can largely ease the process of computing the BBIC parameters for any blocks of interest. Moreover, a fast progressive integral image generation scheme, seamlessly integrated in a streaming-mode macroblock-based video coding system, is proposed. The experimental results show that the proposed technique achieves an average six-fold speedup, in comparison to the traditional computation methods under typical conditions.