Lossy Image Compression using Frequent Pattern Mining based Huffman Encoding

We propose a novel image compression technique which uses lossy and lossless compression techniques. The proposed work uses both lossy algorithms as well as lossless algorithms. Lossy algorithms constitute a modified DCT algorithm and frequent pattern mining along with lossless algorithms which include huffman encoding and decoding, and modified run-length encoding. We introduce a novel parameter( N) for the modified DCT and perform compression using different parameters that affect the size and quality of the resulting image. Simulations of our proposed work indicates significant improvement in the compression ratio and image quality compared to existing alternatives.