【作者】 王丽娟；

【导师】 余宁梅；

【作者基本信息】 西安理工大学 ， 微电子学与固体电子学， 2008， 硕士

【摘要】 为了实现图像的实时压缩传输,本课题组研制了基于图像块动态划分矢量量化算法的图像压缩芯片,该芯片可以实时完成对灰度图像的压缩。但是,现实世界呈现给人们的是丰富多彩的景象,只有彩色图像才能真实地反映客观世界,颜色信息是彩色图像非常重要的视觉特征,因此,单纯灰度图的压缩在实际应用中受到较大限制,研究开发彩色图像压缩系统刻不容缓。本文在现有研究基础上,提出了一种基于PDVQ算法的彩色图像压缩方案。基于已有的灰度图像PDVQ芯片,在FPGA上创建外围电路,连接外部摄像头和PDVQ图像压缩芯片,并把压缩数据传送给PC机;用VC6.0实现接收数据和解码图像的功能,实时地解码和显示图像,最终设计完成了基于PDVQ的彩色图像压缩编解码系统。同时,针对灰度图像压缩码书不能很好的完成彩色图像压缩还原的问题,本文提出了彩色图像矢量量化码书的设计方案,采用经典的LBG码书训练生成方法,设计出适合彩色图像编码的码书。并用设计的彩色图像码书,对常见的几幅彩色标准测试图做了编码实验,进行了性能比较,最终确定出了性能优良、尺寸合适的彩色图像压缩码书。与PDVQ芯片中的灰度图像码书相比,此彩色图像码书的尺寸只是其1/16,还原图像的峰值信噪比(PSNR)平均提高了73.51%,匹配速度也有16倍的提升。此外,本文还讨论了一种基于积分投影的矢量量化算法,并对此算法的性能进行了研究讨论。实验结果证明,与普通VQ相比,此算法的码书存储面积减小了一半,编码速度也有近一倍的提升,而解码图像的峰值信噪比(PSNR)平均仅降低了0.25dB。完成了该算法的电路设计及RTL级代码编写,通过RTL级仿真验证了代码的正确性,最终完成FPGA的实现。同时,通过实验证明基于此算法的彩色图像压缩效果良好。更多还原

【Abstract】 In order to realize images’ real-time compression and transfer, research team invented a real-time image compression chip which based on PDVQ algorithm. The chip can real-time accomplish the compression for gray images. However, the real world presentation format that gives people is colorful scene. Only color images can truly reflect the objective world. Color information is a very important visual feature of color image, so the compression of simple gray images is restricted greatly in actual application. Therefore studying and developing the color image compresses system is extremely urgent.On this basis, this paper proposed a project of color image compression based on PDVQ. Based on the gray image compression PDVQ chip, this paper established the outer-ring circuit on FPGA, linked the outside camera and PDVQ image compression chip, and transmitted the compression data to PC machine. Receive compression data with VC6.0 and decode image, then show the image. Finally, complete a color image compression system based on PDVQ.At the same time, as gray image codebook can not compress the color images very well, this paper proposed a color image vector quantization codebook design plan. Several codebooks which are suitable for the color images coding were designed out using classical LBG codebook generating method. Using these codebooks, several standard color testing pictures were coded, and their functions were compared. Ultimately, high performance and suitable dimension’s color image vector quantization codebook were confirmed. Compared with the gradation image codebook in PDVQ chip, this color image codebook dimension is only 1/16, and the peak signal noise ratio (PSNR) value of decipher image has improved 73.51% averagely, then the matching speed also has lifted 16 times.Besides, this paper proposed and discussed an algorithm of vector quantization based on integral projection. Compared with standard VQ algorithm, the codebook stored area of this algorithm reduces 50%, and the result of experiment shows that the encode speed can be improved almost one times while the ordinary peak signal noise ratio (PSNR) only reduced 0.25 dB averagely. The feasibility of this algorithm is verified, and the algorithm is written in Verilog HDL. The coding is verified by simulation. Finally, this design is implemented in FPGA. At the same time, it is proved that the color image compression based on this algorithm has good effect.更多还原