Extensive Survey on Image Compression using Deep Learning

Nowadays, Data Compression is adopted because it leads to pruning transmission bandwidth, storage hardware and data communication time. Deep Neural Networks (DNN) have confirmed exquisite achievements in several tasks such as Image Classification, Image Processing etc. Excellent performances with speech, audio or image input signals differentiates Convolutional Neural Networks (CNN) from other neural networks. Enhancing the video and lossy image compression visual quality is treated as a serious problem, but the modern advances in computing power along with the accessibility of sizeable training datasets has led to the usage of deep learning networks like CNNs to deal with image processing and image recognition tasks. Some of the image and video compression techniques (lossy as well as lossless techniques) include JPEG, JPEG2000, BPG, GIF, PNG, MPEG etc. Data Compression along with DNN ensures that there is not much loss in data, size is reduced and the accuracy also is preserved. In the following section, a comparison of various papers related to data compression and image compression using Deep Learning Techniques like CNN is provided along with suggestions for topics that can be taken up in the near future.