High and Low Frequency Region Separation Method for Adaptive Image Expansion

Signal expansion has played an important role in multimedia signal compression and analysis. There are a variety of signal expansion methods, including DCT, KLT, Legendre polynomial expansion, wavelet transforms, etc. This study proposes a novel approach based on adaptive region segmentation. The main architecture of the proposed method is to divide images into high-frequency and low-frequency components. Then, these components are further divided according to the connectivity and the region size. Then, the adaptive DCT is applied to expand each part. With the proposed method, the truncation of the high-frequency component in the edge part will not affect the expand result in the low-frequency region. It can effectively reduce the ringing effect and decrease the error of image expansion. Experiments demonstrate that, with the proposed frequency-adaptive segmentation method, the accuracy of image expansion can be much improved. The proposed expansion approach is expected to be applied to other two-dimensional signals.