Image coding method using DCT in skew coordinates

In DCT coding of an image, the image is divided into square blocks, and variable-separation two-dimensional DCT is applied, with one-dimensional DCTs applied to the rows and columns in cascade. In the variable-separation transform, the power packing efficiency is low at the edge in the slant direction of the image. Another problem is that mosquito noise may be produced near the edge in the restored image, which degrades the image quality. This article proposes adaptive DCT coding in which the two-dimensional DCT on the skew coordinates and variable shape region segmentation according to the direction of the edges are introduced. In the proposed method, the shape of the region is restricted to rhombic, triangular, and rectangular blocks to simplify the segmentation and transform processes. The DCT for triangular block is a nonorthogonal two-dimensional DCT in which the normalization factor is determined so that the DC base is used as the transform base. This helps to improve performance in the flat part of the image. The power packing efficiency is estimated, and it is shown that the DCT in skew coordinates can improve the power packing efficiency compared to the conventional square block DCT, and achieves results comparable to the direction-adaptive Karhunen-Loeve transform. The result of a simulation experiment is shown, which shows that the proposed method is useful in reducing mosquito noise.