Counting Cell Using Digital Image Processing Technique December 19, 2005 Prepared By: Pei Qi Yang Wang Electrical&Computer Engineering University of Wisconsin-Madison ECE 533 Digital Image Processing Instructor: Prof. Hu, Yu HenABSTRACT This project is focus on developing a computer based software program to be used by the biomedical department to aid their research need. The biomedical department is conducting a research requires an automatic counting program to count the number of iron contained blood cells from a photonic image. The main objective of this project is to explore and develop an image processing program to accomplish automate counting process. In this project, one of automate counting programs being proposed and developed. The basic idea of this program is to use correlation function to find the largest summation between a sampled image (the wanted cell) and the giving image. Since the largest summation represents the location where the two image shares similarity, it can help us identify the location of wanted cells on a given image. By taking a certain threshold, those wanted cells can be extracted and counted. This project explains how each process was considered and developed. Because this program is intended to be used by researchers of biomedical department, who has limited knowledge in image processing tool, as a developer, we also have to consider a user friendly interface to make our program more useful to the user. Furthermore, our scope might go beyond this project. It is also our interests to use this chance to continue our exploration and probably solve a more general task of object recognition in the future.INTRODUCTION The biomedical department is conducting a research on the study of iron absorption in the blood cells of an orphan. Studying iron absorption requires work in counting number of iron contained blood cells from a photonic image. So far, hand counting is the only method being used. However, due to high vulnerability in human error and large time consumption, better and more effective image processing software is needed. In this project, an automatic counting program is developed, based on image processing techniques. APPROACH & IMPLEMENTATION Since this software will be used by a biomedical researcher, we would like the user to have total control over his or her experiment. First, we would like the user to load an image in the software. Then, it will scan the image and ask the user to select a sample of a specific object of certain color to count with. Based on the range of hue level of user’s selection, the program will search and find all the similar shape and color contained in the image. User can choose to delete or add any object with his or her choice. After the selection is finalized, the program will calculate the total number of wanted object in the image. On the technical side, after an image is loaded and sub-image is selected by the user, first, the program will run an autocorrelation between the image and sub-image. A program fixed threshold is chosen to pick out the largest summation between the two images. By identifying the location of these largest summations, the image can be further changed into binary image to simplify the counting process. Based on the selection, the image is changed into a binary image. Then, the binary image is imposed on the original image so that the user can add or delete any cell of his wish. After the change is finalized, a technique which is similar to the erosion technique is used to change all the wanted shape into a single pixel. The sum of the single pixels will be the number of wanted cell. Based on the presented approach, the implementation basically involves five critical sections: 1) Choosing Sample To a great extent, the result of Counting Cell relies on the people who was doing a count task. Experiential determination always significantly affects the result of counting. Our approach takes full advantage of people’s experience, therefore we firstly let researcher self-select the sample element for counting. There is a branch of factors that could influence the selection of sample, however among those factors, the most obvious, and straightforward criteria is color choosing. Because the blueness of the blood cell illustrate the concentration of iron in the cell, it is hard to decide how blue the color has to be in order to be counted. This criterion should not be decided by the developer, but the lab researcher. This is the reason why we let the user to select a range of blue color for us to use. This feature enables us to apply digital image processing technique to acquire the sample more easily. In our case, HSI color mode firstly was used to extract specific color elements from original image.When humans view a color object, we describe it by its hue, saturation, and brightness. Hue is a color attribute that describe a pure color, whereas saturation gives a measure of the degree to which a pure color is diluted by white light. Brightness (intensity) component is a subjective descriptor that is practically impossible to measure. Among those three components, hue value significantly determines the color scope for any object. Therefore, in our case we extract the wanted color elements through specifying the range of hue value. Considering the range of hue value should vary according to different image, we also allow researchers to adjust the range of hue value to adapt to particular image. 2) Correlation Based on the range of hue value and given shape, the correlation function is used to find similarities between original image and the selected sample. The highest peaks in Figure 1 represent the location of the founded similarities. More over, this peak also reveals where the blue cells are located. The height of the peaks in the correlation figure really shows how similar the two images are at a certain area. Higher the peak represents stronger similarity. Figure 1. Result after applying correlation function on the image. 3) Thresholding A certain threshold needs to be picked to extract the location of the blue cell. At this stage of the program, the threshold is chosen by the program; however, this parameter can be given to the user to decide to obtain a better result. 4) Convert to binary image After a threshold is decided, the program takes any hue values which are above the threshold and assign a RGB to
View Full Document
Unlocking...