Vol.2 No.4 2010

8/68

Research paper : Development of a real-time all-in-focus microscope (K. Ohba)−238−Synthesiology - English edition Vol.2 No.4 (2010) Using this sequential algorithm, sequential processing is conducted while changing the focal distances, and by moving the focal distance to the end, each of the matrixes of the finally updated image memory becomes the all-in-focus image (AIF) and the depth image (DEPTH). While the equation for IQM seems complicated, only Laplacian and smoothing processes are applied as image processing technology. Laplacian is a secondary differentiation, and in the world of digital image, differentiation is the difference from the neighboring pixel, and it becomes Laplacian by differentiating twice. Smoothing is averaging. The two processes are four arithmetic operations, and are optimal for hard logic circuit for high speed. Normally, the memory will hold N frames of differing depth images, the IQM image for each are calculated, and the all-in-focus and depth images are obtained (image memory for total 2N + 2 frames were necessary) by comparing the IQMs at the same pixel position. In contrast, in the sequential algorithm, the obtained images are sequentially compared on the spot, and the memory is needed only for four images consisting of the original image, IQM image, all-in-focus image, and depth image. This reduced the memory requirement, and it became easy to configure the hardware, as it will be explained later.(1) Format init (IQM);(2) Move focal distance for (FV=0 to FVMAX) { (FV: focal distance) mov (FV); move focal distance(3) Image input IO = input; image input(4) Laplacian filter IL = lap (IO); preliminary processing(5) Median filter (average filter) IM = ave (IL); preliminary processing(6) Comparison of brightness value of each pixel, and corresponding copy of image data if (IM(x, y) > IQM(x, y)) { IQM(x, y) = IM(x, y); update assessment value AIF(x, y) = IO(x, y); create all-in-focus image DEPTH(x, y) = FV; create focal distance information } }(7) Image data output output (AIF, DEPTH); image data output4 First FS phaseTo accomplish the above methods and information processing in real time, it is necessary to move the depth of focus at high speed and at the same time, capture and process the images at high speed. Therefore, we determined that the basic system components were the following three:(a) high-speed variable focus mechanism(b) high-speed image capture mechanism, high-speed image processing arithmetic circuit(c) high-speed communication.For example, in conducting the process in eight-step focal distance, to obtain the output in real time at 30 frames/Fig. 4 System configuration for sequential algorithm.Image memory(IO)Image memory(AIF)Image memory(IQM)Image memory(DEPTH)Varifocal lensIntegration unitPreliminary processing unitHigh-speed CCD256*256*8bitsIOIMIM1000fpsNTSC video output30fpsNTSC video output30fpsFVAIFIQMDEPTHDACAmp.Iaplacian filtermedian filterFig. 5 Flowchart for sequential algorithm. If IM(FV,x,y)>IQM(x,y)IQM(x,y)=IM(FV,x,y): AIF(x,y)=ORG(FV,x,y): DEPTH(x,y)=FV: If IM(FV,x,y)>IQM-minYYYORG(FV,x,y)ORG(FV,x,y)IM(FV,x,y) FV=0 FV

元のページ 

10秒後に元のページに移動します