Gradient Feature Based Improved Optimum Non-Negative Integer Bit Allocation for the DCT Based Image Coding
International Journal of Networks and Systems · 2018
The quantization process is the most significant part of any image transform coder, which solely governs the loss characteristics during compression of the images.Hence, several quantization techniques have been developed to achieve better quantization performance for the image transform coders.The Optimum Non-negative Integer Bit Allocation (ONIBA) is one of the impotent techniques that provide optimal quantization bits based on the Variance characteristics of the transform coefficients for the best possible quantization in the image coders.However, the variance feature is not supposed to be the best indicator for all types of variations in the images.As a result, the performance of existing ONIBA techniques can be improved by the utilization of other feature that can accurately estimate the image activities.Therefore, this paper presents a new Gradient Feature-based ONIBA (GFONIBA) algorithm to achieve better quantization for the DCT based image transform coders as compared to the recent ONIBA algorithms.Extensive experiments are carried out to validate the quantization performance of the proposed GFONIBA algorithm.The results show that the proposed GFONIBA algorithm outperforms and provide a significant gain in the reconstructed image quality as compared to the recent ONIBA algorithms.