APBT-based Image Coding at Low Bit Rates with Down-sampling and All Phase DCT Interpolation

摘要

The conventional JPEG standard adopts the Discrete Cosine Transform (DCT), which usually needs the more complex quantization table and adjusting the bit rate results more complex multiplications. Based on the theory of All Phase Biorthogonal Transform (APBT), this paper proposed a novel image coding algorithm based on down-sampling, APBT-based image compression (APBT-JPEG) and interpolation. In the algorithm, down-sampling to the image data after the All Phase DCT (APDCT) filtering before compression and subsequently interpolation to the decompressed image are applied and especially APBT is used in the image compression step instead of the DCT. In the interpolation step, cubic convolution interpolation and the APDCT interpolation were used respectively. Experimental results show that the proposed scheme outperforms the conventional DCT-based JPEG (DCT-JPEG) compression scheme at low bit rates both in terms of PSNR and visual quality.