基于视觉特性选取图像压缩预处理方式

【作者】 李士锋；

【导师】 赵辉；

【作者基本信息】 山东大学 ， 信号与信息处理， 2005， 硕士

【摘要】 就图像传输而言,数字图像处理过程中经常要产生很多包含图像数据的大型文件,而且这种数据会经常在不同的用户及系统之间进行交换。大数据量的图像信息会给存储器的存储容量,通信干线信道的带宽,以及计算机的处理速度增加极大的压力。单纯靠增加存储器容量,提高信道带宽以及计算机的处理速度等方法来解决这个问题是不现实的,这时就要考虑利用图像压缩方法。 很多情况下,我们需要在窄带信道中传输大量静态图像,但并不需要图像多么清晰,而是只需要让人眼能够识别出来即可,这时我们通常选用高倍有损压缩来对图片进行处理。 对于图像的高倍有损压缩来说,面临这么一个问题:压缩率太高时,压缩后的视觉效果往往不能满足人视觉需要,为解决这一问题,可以在压缩前增加预处理环节。本文提出了基于视觉特性选取图像压缩预处理的方式,即,在对图像进行压缩前,根据人的视觉掩盖效应及其采用的压缩方法和压缩后要求的效果不同,选取不同的预处理方法,将其压缩时会产生较大失真的视觉敏感部分做增强处理,减少这部分信息量在压缩时的损失,进而在一定程度上实现图像压缩比的提高和图像质量的明显增强,较好地解决了图像实现高压缩比与图像质量下降过大的矛盾。 选取合理的预处理方法,对小波压缩前的图像进行增强处理,可以有效的抑制图像的视觉不敏感部分,增强视觉敏感信息量在图像中的比重,同时能够使小波变换后系数量化的损失相对较少,解码后恢复图像的PSNR数值保持较高的水平。更重要的一点是:针对图像压缩效果的不同需要,使用相应的图像预处理方法可以在同样的压缩率下提高图像视觉质量,或者是在相近的图像质量下实现图像的压缩率的提高。更多还原

【Abstract】 As far as image transmission is concerned many large-scale files including image data will be produced in the digital image processing. Furthermore these data are often exchanged among different users and systems. Large image information bring much compressive stress to capacity of memorizer, channel bandwidth of communications trunk and processing speed of the computer. It is not practical to improve hardware performance merely. Here we take image compression into account.Usually a mass of static image transmitted in narrowband channel need not greatly clear-cut but legible to HVS (human visual system). Thus we choose image processing of large compression rate.The problem is image quality not good enough to satisfy with HVS in processing of large compression rate. So we add a step before image compression, pretreatment. This paper present a pretreatment method of narrowband image transport based on HVS. That is to say, we choose different pretreatment according to different compression format with visual covering effect. The sensitive image part of HVS that will greatly distort is be enhanced to reduce the losing information in image compression. Consequently the pretreatment method realizes enhancement of the compression ratio and improvement of image quality to a great extent so that to clear up the conflict between large compression and greatly digressive image quality.If reasonably choose the pretreatment method to make enhancement processing to original image before wavelet compression, we can effectively restrain the insensitive image part of HVS, increase the proportion of sensitive information to the whole image, and reduce the losing of coefficient quantification in wavelet transform so that the PSNR value of resumed image keeps relatively high level in decoder. More importantly, if reasonably choose the corresponding pretreatment method based on need of different image compression formats, we can improve visual quality with the same compression ratio or enhance compression ratio with similar image quality.更多还原