Bandwidth-aware video encoding with adaptive image scaling

We proposed an adaptive MPEG-2/4 video encoding scheme suited for a dynamic networking environment. Conventional approaches use large quantization scales to reduce the transmission bit rate in low bandwidth situations. However, the larger quantization scale degrades the image quality severely. We believe not only that adaptive quantization should be used for bit allocation, but also that the image resolution should be jointly adjustable. Our goal is thus to investigate how image subsampling should be used in bandwidth fluctuations (e.g., 3/4G wireless networks). For this purpose, a rate-distortion model considering image subsampling is studied and implemented in the proposed encoder. Experiments show that, by choosing the right subsampling factor, a better image quality with considerably lower transmission bit rate can be obtained. A maximum gain in PSNR of 2 dB is observed with more than 50% reduction in quantization size.