Image coding by the mesh block transform

A novel method for image coding, the mesh block transform (MBT), is proposed. It aims at reducing the well-known blocking effects, artifacts very common in transform coding schemes. The approach used is to divide an image of dimensions MN×MN into M² meshes of dimensions N×N. Adjacent pixels in a mesh are separated by N pixels in the image. Each mesh is transformed and the same frequency coefficients from all meshes are grouped into blocks. A second transform is then applied to these blocks and its coefficients are quantized. A mathematical analysis of the MBT is done and a coder is proposed. The performance of the method is evaluated. It is concluded that, after some improvements, the MBT may be a good alternative to the classical transform coding schemes, especially when highly detailed images are involved