A Study on a New Principle of Image Compression

研究取得的主要成果如下。(i)为了更好地逼近真实的图像信号的幅值特性，提出了一种新的图像生成模型--修正AR模型。(ii)传统的编码方案，如DCT、DPCM和分形编码，对通过使用（i）产生的伪图像信号显示出与对真实的图像信号几乎相同的压缩性能。(iii)提出了一种产生混合图像信号的频谱交换方法，将真实的图像信号的幅度谱或相位谱交换为伪图像信号的幅度谱或相位谱。(iv)用（iii）的技术产生的各种混合图像信号的压缩实验表明，对于DCT、DPCM和分形编码的压缩，仅使用幅度谱的信息而不使用相位信息。(v)很清楚，表征图像的形状信息被包括在传统编码方案中不使用的相位谱中。(vi)强烈希望将相位信息即形状信息的直接使用引入图像压缩方案中。(vii)作为直接利用形状信息的图像压缩方案之一，提出了一种新的可变形状分形编码方案，该方案将小的基本块整合成一个与形状相适应的范围块。（viii）上述方案通过更有效和精确地近似形状来实现压缩性能的改进。(ix)作为有效利用形状信息的第二种方案，提出了任意形状DCT方案。(x)基于DCT的任意形状编码方案大大提高了压缩性能。

Main results obtained in the research are listed below.(i) A new image generation model, Modified AR model, was developed to approximate the amplitude characteristic of real image signals well.(ii) Conventional coding schemes such as DCT,DPCM,and fractal coding show almost the same compression performance to pseudo-image signals generated by using (i) as to the real image signals.(iii) A spectrum exchange method to generate hybrid image signals was proposed, where amplitude spectrum or phase spectrum of the real image signal is exchanged to that of the pseudo-image signal.(iv) Compression experiment of a variety of hybrid image signals generated with the technique of (iii) shows that only the information of the amplitude spectrum is used and phase information is not used for compression by DCT,DPCM,and fractal coding.(v) It is made clear that shape information which characterizes an image is included in the phase spectrum that is not used in conventional coding schemes.(vi) It is strongly desired to introduce a direct use of the phase information, that is, shape information into image compression schemes.(vii) As one of the image compression schemes to use shape information directly, a new, variable shape fractal coding scheme that integrates small basic blocks to form a range block adjusted to the shape was proposed.(viii) The above scheme achieves the improvement of compression performance by approximating the shape more efficiently and accurately.(ix) As the second scheme that uses the shape information effectively, an arbitrary shape DCT scheme is proposed.(x) A great improvement of compression performance is obtained by the DCT-based arbitrary shape coding scheme.