Research Of Cell Image’s Preprocessing And Segmentation Based On DSP

In the automatic microscopic examination system,the real time and accuracy ofdigital image processing technology are more and more important. Theimage-processing system based on DSP has significant advantages and now plays animportant role in the field of embedded image-processing; Then the preprocess andsegmentation plays a fundamental role and lead to success or failure in the system.Around the microscopy system of AVE, this paper did research on the cell imagepreprocess and segmentation, combining the TMS320C6678LE board, did researchon how to the realize and optimize the edge detection, the edge processing and theextracting target algorithm on DSP.First, statisticed and analysised a large number of cell image characteristics.Based on wavelet transform, revolves the two significantly cell image characteristicsproblem. For the noise, adoped an adaptive threshold to denoise in the high-frequencyand it works well. For the shadow, based on regional characteristics, proposed athreshold method in the low-frequency for shadow elimination.Secondly based on the Canny operator what is a classical edge detectionoperator, transplanted to the DSP. And Combined the steps of the Canny such asgaussian filter, histogram statistics, difference image, grad vector calculation with theDSP optimization techniques and method such as packaging data, wide memoryaccess, reduce storage dependence, manually written parallel code. Realized anefficient operation of the Canny on DSP.Then aiming at the shortcomings of the Canny operator for edge detection, putforward morphology processing with multiple structure for the main problems. As themorphological dilation and erosion algorithm were frequently used, proposed amethod to improvement the morphological process, which led the morphologicaloperation to memory move and bit operation on DSP.Finally chose the inner boundary and chain code to express the cell image target.proposed that using region growing to get the image’s closed circular edge. Combinedwith the DSP, researched the pass to non-recursive algorithm. And came to aconclusion that try to avoid using recursive algorithm on DSP.