DEVELOPMENT OF A HYBRID BTC ALGORITHM USING DEEP LEARNING

The article proposes a new hybrid approach for Block Truncation Coding (BTC) that uses deep learning methods. The suggested method predicts the best threshold values for each image block using a deep neural network, and then applies those values to represent the image with less bits. To maximize the compression performance, the system is trained on an image dataset using a mix of supervised and unsupervised learning techniques. The results of the experiments demonstrate that the suggested algorithm works better in terms of compression ratio and image quality than conventional BTC approaches. Additionally, the technique can withstand a variety of image sizes and kinds, which makes it appropriate for a wide range of image compression uses. The suggested method presents a viable strategy for enhancing BTC's performance and expanding the picture compression field.