Low bitrate video coding using wavelet transform

In this paper we present a method for video coding which consists of wavelet transformation, motion estimation and context based quantization and coding. The wavelet transformation is realized using a lifting scheme for biorthogonal filter bank 9/7. The motion estimation is based on an edge detection algorithm which is preformed in the wavelet domain e.g. on approximation at low frequencies and details at high frequencies. The edges are assumed as points where motion is detected. The search area for a block matching algorithm is selected around edges in the frame. The wavelet coefficients, motion estimated frames and motion vectors are quantized using a context modeling of quantized coefficients and a context based entropy coding method. The method for video coding is compared using a PSNR measurement and a target bit-rate with a TML-8.4 method using a standard video test sequences. The proposed method is competitive with the TML-version 8.4 in a PSNR measure. The paper ends with measurement of computational complexity of the proposed method and a conclusion.