A Scalable Video Coding Scheme Using KL Transform in Multiple Linear Subspaces
基于多线性子空间KL变换的可伸缩视频压缩方法

This paper proposes a novel transform coding method based on the Karhunen-Loeve (KL) transform in multiple linear spaces. This method uses multiple linear subspaces to approximate image signals, and use a KL transform for each linear subspace, so that the transform coefficients in each subspace is close to Gaussian, which effectively de-correlates the coefficients. An SNR-scalable video coding scheme is developed by combining the method with wavelet decomposition. Experiments prove that this scheme accomplishes better coding efficiency than the schemes employing DCT, DWT, and Directional DCT methods.