相关反馈方法是对基于内容图像检索系统的有效改进，它将人类视觉特性逐步引入检索过程，有效地减小了图像低层特征表示与图像语义理解之问的差异。但传统的相关反馈算法存在反馈次数多，且无法积累用户反馈信息等缺点。本文针对这些缺点，在相关反馈图像检索系统中引入了可更新特征库。即在原始特征索引库的基础之上引入了一个用户可修改的特征索引库，系统可以将用户多次反馈的信息逐步嵌入到这个特征索引库中。与此同时，我们提出了一种基于多分辨率分析的彩色图像纹理特征描述方法，并将其用于特征库的构建中。我们在一个含有10 000幅图像的图像库上所做的测试结果表明：与Illinois大学的MARS系统相比[1]，本系统可明显提高系统检索准确率和相关反馈的收敛速度。

A new image data compression method using fractals and the discrete cosine transforms (DCT) is presented. The original image is first encoded by fractals in the DCT domain, then the error image is encoded using the DCT. Experiments show that the method can achieve high fidelity at a high compression ratio.

The fractal image coding technique has attracted a degree of interest for its low bit rate. But the reconstructed image is of medium quality. This problem has prevented the fractal technique from being practically used. In order to improve the compression fidelity, a new affine ransformation is roposed in this paper. Meanwhile, its contractivity requirement is analyzed, and the optimal parameters are derived using the least square method. The new affine transformation has been practically used in image coding. Experiments show that the PSNR can reach 28.7dB at a compression ratio (CR) of 16.4 for the 256 256 8 "Lena" image. Comparison with other fractal coding schemes shows that the new affine transformation can improve the reconstructed image quality efficiently.

Multiple description (MD) coding has many applications in transmission of images over unreliable packet or multiple path networks that can not guarantee lossless data delivery. In this paper, an effective multiple description image coding scheme is introduced. This scheme is mainly based on the wavelet transform and multiple description lattice vector quantization (MDLVQ). The haracteristics of wavelet coefficients in different frequency subbands are taken into account in the design of the MD image coder with different optimized MDLVQ parameters. The experimental results are presented to demonstrate the effectiveness of the proposed scheme.

Abstract. In conventional fractal image coding (FIC) schemes, domain blocks are constrained to be twice as large as range blocks to ensure the convergence of their iterative decoding stage. However, this constraint has limited the fractal encoder to exploit the self-similarity at the same resolution scale of natural images. To overcome the shortcoming, a novel scheme using same-sized range and domain blocks is proposed. Further, a recursive scheme feeding the coding results back to the input during the encoding procedure is used to improve the decoded image quality. Experimental results show our method gives significant improvement over Fisher's FIC.