Research On Lossless Compression For Natural Color Images

With the rapid development of information technology, digital images as an intu-itive and vivid information carrier has been used more and more frequently and broadly,which results in the problems of transmission and storage about digital images. Uncom-pressed images require larger memory to store and larger bandwidth when transmitting.Image compression technology tries to mitigate these problems via reducing the data of original images with only a few information loss or without information loss, thus has at-tracted many attentions from researchers. In many important situations such as military,law, medical, remote sensing and art works protection fields, image quality degradation is not allowed, these images have to be compressed via lossless compression methods.It can be seen that there is a need to research lossless color image compression methods.Our research starts with the formation process of natural color images in digital cameras, then we analysed the redundancy features of natural color images carefully.Based on these features, we proposed a new lossless compression algorithm for natural color image. At first, the color images are transformed to a new color space from orig-inal RGB color space via color space transformation. In the new color space, the new color components are less correlated. In the next step, we generated the improved Lapla-cian Pyramid representation for every color component in new color space. In this way,every color component can be represented by a down-sampled image and several resid-ual images. In the last step, entropy encoding method such as huffman coding is used to coding these residual images and down-sampled image of every color component. In the process of generating the Laplacian Pyramid representation of images, image interpola-tion method is used to magnify images, the better image interpolation method performs,the better compression performance can be achieved. So as to choose a proper image interpolation method, we proposed two simple image interpolation methods and com-pared them with five traditional image interpolation methods. At last we implemented the whole lossless compression algorithm in C++, and compared with commonly used methods as TIFF, PNG and CALIC. Experimental results show that, on the average,the proposed algorithm improves 26.95%, 19.01%, and 10.43% over methods used in TIFF, PNG image format and CALIC algorithm respectively, which means the proposed algorithm is more effective.