A method of approximately reconstructing JPEG quantization table via saturated overflow

Header information is vital to decode and display a JPEG file. Therefore, it is necessary for head-missing JPEG fragment to estimate the header information. In this paper, a novel method is proposed to approximately reconstruct JPEG quantization tables. Based on the saturated overflow phenomenon introduced in the inverse DCT transform, the original quantization tables can be approximately rebuilt by our algorithm. The experimental results show that the estimated quantization tables can be used to normally display the image with no visual distorted. Experiments also prove that the method is more effective when the compression quality is high.