Digital Microscope Image Sharpness Determination Method

The sharpness of the image needs to be determined and compared in theparfocality detection of digital microscope lens. However the result of the imagesharpness metric is highly affected by the light factors which lead to unstabletest result. So it is necessary to research on the methods that have strongrobustness to illumination for determining the sharpness of the image.In this paper, we analyze and compare the common methods for imagesharpness metric, aiming at illumination which affects at method for the digitalmicroscopic image sharpness metric, designing three kinds of methods whichare robust to illumination in image sharpness metric.Local normalization is a common method to process illumination. Weapply it to process image, which eliminates the effects caused by light factorsand gets a normalized image which is not sensitive to illumination. Then use thecommon image sharpness metric methods to determine the sharpness. Theexperiment results show that this method has some robustness to illumination,and the stability can be improved by76.05%averagely and90.22%maximum.The image texture is an important index for image sharpness metric. Thispaper adopts the local binary patterns which is capable to describe texturefeature excellently to extract image texture feature. On this basis, accord tobinary number connection fluidness, different target images adopt different wayto connect, and get the best LBP image, finally determine the sharpness of LBPimage. The experiment results show that it can determine the sharpness of thenon-uniform illumination image excellently based on the texture feature. Thestability can be improved by83.89%averagely and99.90%maximum.Uneven illumination is a common light problem. We apply oriented localhistogram equalization method which has the image enhancement effect todetermine the sharpness of image. Combining with its edge directioninformation, adopt different edge direction according to different regions toobtain the best enhanced image, finally determining the sharpness of the image. Through a variety of methods test, this method has a good robustness toillumination. The stability can be improved by87.66%averagely and99.28%maximum.