The Design Of Microscope Image Processing And Analyzing System Based On CMOS Sensor

Microscope is very important in researching and studying as a auxiliary tool. In recent years there is a dramatic development in image sensors such as CCD and CMOS, and in various image processing algorithms. So microscope is heading for digitization and automation day by day.A micro-image processing & analyzing system is built in this article, which select CMOS as a image sensor. The article includes system introduction in whole scale, working principle of the hardware, image correcting-processing pipeline and several important micro-image analyzing algorithms. The latter two make the main parts of this article.As is known, image color correction is an important aspect in digital image processing. But the mainstream is used in digital cameras, and there is few application in micro-image. Based on the exposure theory of CMOS and the chromaticity theory, and considering the special working conditions, an image processing pipeline is introduced.The pipeline includes several main steps such as linearization, color correction, white balance, images enhancement and so on. New methods proposed in those aspects are more suitable than the traditional.A non-linear white balance method is proposed in this paper for the first time. Taking XYZ stimulus and RGBs of each color block in Munsell color chart illuminated by D65 light source as training samples. A BP neural network mapping relation is built between digit-image RGB values of standard Munsell color chart and XYZ stimulus. Then Estevens matrix is introduced, which is used in translating XYZ stimulus to cone stimulus. Then use a non-linear calculation, at last translate back into digit-image RGBs, and an image after white balance is produced.With the miniaturization trend of CMOS chips and individual pixel, a problem appear known as hot pixels and bad pixels, which result in the loss in photosensitive properties. In bad pixels diminishing, in this article an EEPROM is proposed which save bad pixels coordinates obtained in long exposure condition; when the lighting condition is normal, query bad pixels information and diminish.In non-linear correction, a method is introduced that taking images of arithmetic gray card in constant lighting condition, then a look-up-table is built to correct the non-linear phenomenon of CMOS chips. In this way, the linearization performance of CMOS chips improved greatly.As shown in the results, the video-processing output following the pipeline proposed in this article is more realistic.Another main aspect is that several important image analyzing algorithms are proposed, including image geometric feature deep analysis, and optical density measurement. As a complement component, a hardness measurement algorithm is proposed.