Improved Image Compression using Autoencoder and Discrete Cosine Transformation

This study proposes a method for improved image compression using a combination of discrete cosine transformation (DCT) and autoencoder. Images typically contain large amounts of data that require significant storage space, making them difficult to store and transmit. Compressing images is a practical solution to this issue as it reduces memory usage and enables faster transmission to the receiver. In this approach, we use DCT as a preprocessing step before training an autoencoder model to compress the image while retaining all essential information. The proposed method involves a convolutional neural network (CNN) that performs down-sampling and up-sampling operations on the input data processed by DCT. The performance of the proposed method is evaluated and compared with traditional image compression techniques such as JPEG, JPEG 2000, and BPG. Experimental results demonstrate that the proposed approach outperforms the traditional techniques in terms of compression ratio and image quality.