基于改进的正交FRIT的分层图像编码算法

Multi-layered image coding based on modified orthonormal FRIT

针对直线边缘较丰富的图像,结合小波变换和脊波变换的分析优势,提出了一种新的分层图像编码算法。本算法首先对原图像进行一级小波变换,取出低频子带LL1进行小波零树编码,从而编码图像的平滑层;然后根据残差边缘层有限Radon变换系数的特征,提出了一种改进的正交有限脊波变换,用以对残差边缘层进行更有效的稀疏表示,并根据变换系数的特点,改进零树编码算法SPIHT,实现了变换系数的渐进编码。实验结果表明,本算法在码率较低的情况下具有优于小波零树编码算法的压缩性能,并且能够更清晰地重建原图像中的直线边缘。

This paper proposes a new multi-layered image coding algorithm used to compress straight-edgerich images. This algorithm combines the advantages of wavelet and ridgelet transformations. After onelevel DWT, wavelet-based SPIHT is used to code the low frequency sub-band LL1 so that the smooth layer of the original image is coded. According to the characteristic of the residual image＇s FRAT coefficients, a modified orthonormal FRIT is proposed to represent the residual edges more efficiently. Then SPIHT is also modified to realize the successive coding of the transform coefficients. Experimental results show that at low bit rate the proposed algorithm is more efficient than the wavelet-based SPIHT and can reconstruct the straight edges more clearly.