Referenceless Full Reference Image Quality Metric Estimation For Embedded Codec
Md Amir Baig, Athar Ali Moinuddin, Ekram Khan · 2019 International Conference on Computing, Power and Communication Technologies (GUCON) · 2019
The quality scalability feature of embedded codecs allows the decoding of image of different qualities from same fully encoded embedded bitstream. In this paper, this feature is exploited to estimate the full reference image quality assessment (FR-IQA) metrics of decoded images without reference image for SPIHT coded images. Any FR-IQA metric either measures similarity or dissimilarity comparison of the distorted/decoded image with respect to the reference image. The proposed work is based on the fact that if the similarity between the reference image and a lower quality image is known, the similarity of images of quality in between can be estimated without having full embedded bitstream or reference image. The proposed method utilizes machine learning regression algorithm for training and testing purposes. The estimated FR-IQA metrics are remarkably accurate. Moreover, each of these metrics require computation of only one corresponding FR-IQA metric as a feature, therefore it is suitable to estimate the FR-IQA metrics of reconstructed images from an embedded codec at the expense of same time complexity as of FR-IQA metrics of the choice. The FR-IQA metrics are usually accurate and fast and therefore, proposed algorithm can be used in real time applications.