International Conference on Computer Engineering and Technology, 3rd (ICCET 2011)
47 Estimation of Optimum Number of Eigen Vectors for Compression of Color Images Using KL Transform
Download citation file:
- Ris (Zotero)
- Reference Manager
The Karhunen-loeve transform expresses image data in terms of number of eigen vectors depending upon the dimension of an image. In color images, there is a large data redundancy available. Thus there is a scope for redundancy minimization and subsequent image data compression. This paper presents a technique to estimate the optimum number of the eigen vectors required for reconstructing the original image without a visible compromise on image quality. The MSE and PSNR are used as benchmarking parameters for reconstruction while 2% loss of energy is considered as threshold for the reconstruction as human visual system can not respond less than 2% loss of energy incident on human retina. This experiment was carried out on many real life images and it has been observed that around 20% eigen vectors of the total vectors are enough to reconstruct original image with more than 40db PSNR in most of the database images. The technique is found to reduce the number of eigen vectors required for reconstruction with 40 db PSNR, to less than 50% in case of most of the real life color images and specially for video scenes where there is a minimal difference between the successive frames of a video scene.