Research On Microscopic Image Fusion

As an important field in image fusion, microscopic image fusion has been applied widely in the field of biomedicine, microelectronic product visual inspection and material inspection and so on. It is well known that optical lenses of microscope normally have high resolving power and suffer from the problem of limited depth of focus. Sinceonly a small part of each image slice is in focus, the visibility of the objects in a single image is limited, even for relatively flat specimens with 3D structures. Microscopic image fusion is used to produce a completely clear image by fusing the registered images of the same scene with different focus, which are under the same imaging conditions.There are many kinds of microscopic image fusion algorithms, first, two algorithms based on spatial area and the frequency area are introduced in this paper with a simple comparision. Next, the microscopic image fusion based on area wavelet transform is discussed extensively, particularlly the fusion rules of wavelet coefficients. After many fusion experimens, the microscopic image fusion rule of low-frequency based on edging and high-frequency based on object clearness and object blur judging was used to select wavelet coefficients. The fusion effect was assessed by emplouing subjective estimate criterion and objective estimate criterion, with the result indicating that the algorithm produce better fusion effect. Then, to avoid the false color being produced after the color microscopic image fusing, the algorithm of fusion rule which is region contrast with proportion and region object clearness and object blur judging was used based on wavelet decomposition without down sampling, the algorithm of fusion rule was used to confirm which pixel is clear pixel, then select clear pixel from the source microscopic image directly. The experimental result shows that the algorithm produces better color microscopic image fusion. Finally, due to the small vision field of microscope, only some part of the image can be gathered once a time, so large field of vision of image is hard to collect. In order to observe or measure the big object, need mosaic many overlapping image to make them an integrative image. Therefore, the key technologies of microscopic image mosaic and correlative mosaic algorithms are introduced in the paper, and the algorithm of template matching based on wavelet transform is used to find the position which is match. The experimental results show that this mosaic algorithm has the advantage of high speed and exactness.