Title:JQTS: Optimal JPEG Quantization Table Selection by Deep CNN-based Texture Prediction

Abstract:JPEG has been a widely used lossy image compression codec for nearly three decades. The JPEG standard allows customized quantization table to be used; however, it's still a challenging problem to find an optimal quantization table with acceptable computational cost. This work tries to solve the dilemma of balancing between computational cost and image specific optimality by introducing a new concept of texture mosaic images. Instead of optimizing a single image or a set of representative images, the simulated annealing technique is applied to texture mosaic images to search for an optimal quantization table for each texture category. A deep CNN model is used to learn those texture features and predict the new image's texture distribution, then aggregate texture optimal tables to come out an image specific optimal quantization table. On the LIVE database, our experiments show a size reduction of 22.46% compared to the JPEG standard table with a slightly 0.3% decrease in the FSIM score. Our method also achieves a further 6.16% JPEG size-reduction against prior work while enhancing the average SSIM and FSIM by 0.24% and 0.07%, respectively.