The Optimal Quantization Matrices for Jpeg Image Compression from Psychovisual Threshold

The JPEG image compression method has been widely implemented in digital camera devices. The quantization process plays a primary role in JPEG image compression. The quantization process is used to determine the visibility threshold of the human visual system. The quantization tables are generated from a series psychovisual experiments from several angle points of experimental views. This paper proposes psychovisual threshold through quantitative experiments for JPEG image compression. This experiment investigates the psychovisual threshold based on the contribution of DCT coefficients on each frequency order to the reconstruction error. The average reconstruction error from incrementing DCT coefficient is investigated to produce a primitive psychovisual threshold. The psychovisual threshold is designed to give an optimal balance between quality of image reconstruction and compression rates. A psychovisual threshold is obtained to generate new quantization tables for JPEG image compression. The performance of new quantization tables from the psychovisual threshold are analyzed and compared to the existing default JPEG quantization tables. The experimental results show that the new quantization tables from the psychovisual threshold produce higher quality of image reconstruction at lower average bit-length of Huffman code than default JPEG quantization tables.