Interpolative Model on Hueristic Projection Transform for Image Compression in Cloud Services
Keywords:
Image Compression (IC), Dense neural Net (DNN), multi-layer perceptron (MLP), Convolutional Neural Net (CNN), Deep Neural Net (DNN), Machine Learning (ML), Deep learning (DL), Orthogonal Transforms (OT),Heuristic Projection Orthogonal Transform (HPOT).Abstract
Image compression analysis have become most prominent feature to realizeto observe, process and store image and its features in local or cloud storages. Since the trend of ML and DL have become the important features that are effective in real time design, we tend to observe the loss and its data manipulation via manual or automatise the compression standard based on the applications. Since, storing of data in less memory stricture and its architecture have become the recent interests of the compression standard with its time features. Presently, we have observed most of the design are either hybrid or novel structures with its transformative analysis to reduce the performance parametric with less memory storage and also less time consumption. But the prominent change of the Image compression rate with its transformative approaches have depicted more loss in image while compression ration increases with pixel quality. So, as to provide such change in the image compression our proposed model implemented a novel intuitive orthogonal Transformation and Heuristic Projection Orthogonal Transformindicating loss less than 1% observed with HPOT-Dense-NN. The effective parametric with different image dataset such as CIFAR-10, MNIST and 100 sample real time datasets are considered to implicate the overall comparison with State of Arts (JPEG, JPEG-2000, GAN, LSTM and CNN).The overall bitrate, compression rate and PSNR with SSIM are implemented with the proposed design algorithms which have shown improved values better than SOA as tabulated.
