지그비 환경에서 개선된 문턱값과 쿼드트리 분해 및 웨이브렛을 이용한 영상압축 및 잡음제거

Abstract

In this paper, a method of the digital image denoising and compression is discussed, which should be necessary to send a digital image through the ZigBee based networks. The information of digital image may be demanded on more variable application at the ubiquitous society, recently. Therefore, it provides a wide activity for the information of digital image to be acquired, sent, and stored through not only wire but also wireless networks. Also, it is expected to become a vast application such as tele-image monitoring system in sending a digital image through ZigBee networks what is called ubiquitous sensor networks. ZigBee is designed to minimize the consumption power of sensors. So, the lifetime can be a few years with one battery. Also, because the architecture of Zigbee is simple, it has the benefit that its size is small, it is economical and it is designed easily. However, it has the limit that it is difficult to send a image data, because of the low speed, compared with other wireless communication methods. Therefore, in order to send a digital image effectively under the environment of the low speed networks such as ZigBee, it should be necessary to denoise and compress the digital image. In the research of human vision system, it is the well known fact that the eye filters a image into multiple bands. The problem of degradation in high compression of image can be solved to eliminate the subband image which is not sensitive in the human vision among the multiple bands. By using the wavelet transformation which is the image decomposition method similar to human vision system, the original image can be decomposed into some subband images, and then, it can be possible to compress the image effectively considering the importance of each bands. Hence, the wavelet transformation is very effective for image compression, which is similar to human vision system, and provides both frequency information and spacial information of image. The research is actively performing in which many coding algorithm is applied with multi-resolution decomposed image by wavelet transformation, recently. In order to send a digital image under the environment of ZigBee networks, it is important task to minimize a loss of information caused by degradation and to reduce the size of data by maximally lowering the bitrate, though the permissible rate of compression is decided by the kind of application. The experiments is the example of actually implementable tele-monitoring system sending a digital image through ZigBee networks. For this experimental environment, tele-monitoring system through ZigBee networks is constructed virtually. And the Gaussian noise and impulse noise is reduced which can be present by processing the system. Then the compression method of denoising image data is discussed. First, after degradation capturing image with noise is transformed by wavelet, the method of selecting a threshold value is proposed, in which the noise corrupted coefficients is found. The selected threshold value can discriminate between the edge and noise. Second, in this paper, the image compression method is proposed by applying a block quantization based quadtree for the multiresolution decomposed digital image through wavelet transformation.

After the discussion of image denoising and compression, each proposed method is integrated by setting a compression rate control threshold value into denoising adaptive threshold value. So the integrated method can improve the performance of image denosing and compression. Finally, in order to show a validity of proposed method, the experimental results is evaluated by performing image denoising and compression both with proposed method and conventional method.

In the experiments, the image size is 256 x 256 pixel and one pixel has a 8 bit data length. And the number of image is capturing image 4 and general standard image 1. In each image, Gaussian noise and impulse noise is added artificially. For the purpose of evaluating the precise performance of each method, PSNR(Peak Signal-to-Noise Ratio), the objective evaluation measure of comparing the picture quality, and the subjective comparing method are introduced. Especially, PSNR of differential image also introduced by which the effect of nosie reduction and the degree of edge components preservation are quantified.