A Novel SPECK Algorithm for Faster Image Compression

In this piece of submission, we have divided our work into two phases. In the first part, we have proposed a 2 channel perfect reconstruction (PR) Quadrature Mirror Filter (QMF) bank which is to be used as a wavelet filter. The second phase of our work focused on a novel preprocessing technique to be applicable for the reduction of computational complexity as well as to make the codec memory efficient. The proposed filter is utilized through the dyadic decomposition with the addition of pre-processed stage for the SPECK algorithm for compression of standard test images. The said modification provides significantly higher PSNR. The proposed pre-processing scheme, called Zero-Shifting, brings the DC values in similar signed integer range as the AC values, so that the transformed coefficient calculation becomes more consistent. Our experiments demonstrate a significant improvement in encoding and decoding time by making the codec faster than the original one.