Rate prediction for image compression based on lapped biorthogonal transform
LV Yan, Song Li, Xin Shou Tian · 2017
Predicting rate without coding is useful for compressing images in a limited communication network. In this paper, a rate prediction model is proposed for image compression methods based on Lapped Biorthogonal Transform (LBT), such as JPEG XR coding standard. The 3-D mapping relationship among the quantization parameter, the image activity, and the coding rate is constructed. Then the coding rate could be estimated by the image activity. In addition, the quantization parameter could also be used for predicting the compression quality (measured by PSNR). Experimental results show that the relative average error of the predicted code rate is 0.03b/p, which can meet the demand of most common applications. The algorithm is simple and the memory consumption is also low. Therefore, it is valuable for image compression methods based on LBT.