Research On Pmmw Image Fusion Algorithm Based On Multi Scale Transform

The difference of millimeter energy radiated by objects is detected by Passive Millimeter Wave(PMMW) imaging system, to provide PMMW images in which the metal targets can be distinguished from the background. The PMMW imaging system has the advantage of penetration ability without radiation, and it’s hard to be detected by enermy. Because of the limit of hardware, the PMMW images have low resolution and signal to noise ratio. The information contained in PMMW images is single and cannot fit the human visual habits. By combining the PMMW and visual images, the multi-source image fusion technic is able to generate a composite image, in which the imformation is expanded, of the same target and scene.This dissertation depends on a reaserch project. An image fusion algorithm of PMMW and visual images based on Non-Subsampled Contourlet Transform(NSCT) and region segementation, as well as a pretreatment algorithm of PMMW images is proposed. In addition, a multi-frame image fusion algorithm and its hardware designing, which aims at real-time imaging system, is presented.1. The PMMW imaging and multi-source image fusion theories are studied. By analyzing the imaging mechanism and characteriscs, the key of fusion PMMW and visual images is discussed.2. A PMMW image pre-processing algorithm including image denoise and enhancement is proposed. To remove the impulse noise, the noise model is analysed and modified so that the detection of noise is more accurate. Then the Gaussian filtering window is set to be adaptively changed. The image enhancement is conducted by using Laplacian pyramid decomposition to enhance the high-pass subbands and low-pass subbands by Laplacian operater and binary image, respectively.3. The image segmentation algorithm based on Mean Shift is studied. Target areas in the PMMW images are extracted by turning the convengence image obtained by Mean Shift process to a bainary image, so that the target and background is separated.4. Based on the study of NSCT image decomposition and Intensity Hue Saturation(IHS) color space transform, a gray image fusion as well as a color image fusion algorithm is proposed. The high-pass subbands of NSCT are fused by modified first order difference square sum based rule, while the low-pass subbands are fused based on region energy. In addition, an image fusion algorithm based on NSCT and region detection is proposed to suit the application of concealed object detection. The Region of Interest(ROI) is extracted by image segmentation and imaging prior knowledge. The fusion is then conducted only in ROI before its embedded back into the visual image.5. To satisfy the needs of project system, a multi-frame image fusion algorithm, which is faster, is prosposed using the range of target’s movement between neighbor frames. The hared ware of image fusion in the project system is also designed.The algorithms proposed in this dissertation are simulated using real PMMW images on Matlab. The simulate results demonstrate the reliability and effectivity of these algortihms.