An Efficient Implementation of Hierarchical Image Coding

ABSTRACT The third issues are compatibility and scalability. Different bitstreams, having multiple resolution versions of the image/video, are more desirable to different users with various computational capabilities. An efficient technique for hierarchical image coding is proposed, which divides an image into its multiple resolution versions with minimum hardware cost. Since the proposed technique uses discrete cosine transform (DCT) over each frequency band data, obtained directly from DCT domain by multiplying pre-calculated matrices, compatibility can be preserved easily with DCT-based image/video standards. First, we present the 1-D case, followed by an extension to the 2-D case. The proposed coder is observed to result in significant reduction of memory and computational complexity requirement with higher peak signal to noise ratio (PSNR), when compared with traditional hierarchical image coding method. Among many image/video coding technologies, layered coding is a widely used scheme [5]. This coding provides scalability in the quality and size of the reconstructed image/video and supports error robustness when low frequency band of image data is transmitted using a less noisy channel or protected using a more sophisticated channel coder. In the event of the loss in the high frequency band of image data, the low frequency part can be used to reconstruct an image with reduced but still acceptable quality. Therefore layered coding has many useful features for mobile multimedia communications. The main disadvantages are high computational complexity and memory requirement, because filtering operation needs to be done in the spatial domain.