Research About Extracting Leukocyte Areas Based On Improved ITTI And Particle Swarm Optimization

Image region of interest(ROI) refers that the observation of the human visual system under the complex scenes, the visual attention involuntarily is placed to a few objects, and priority will be given for the objective area, this area contains the important image information. The image with only having of ROI, can reduces the amount of computation of process image, and improves the effect of the computing. This paper mainly focuses on the visual attention model about extracting the ROI, and analyzes the ITTI visual model for extracting the Leukocyte areas in bone marrow cells image. Currently our medical checking level is not yet higher intelligence. Pathological analysis shows that, in the process of checking bone marrow cells image, human eye focuses mainly Leukocyte areas, and then identifies the anomaly. The manual mode of operation inevitably arises the worker's fatigue,and further led to drop minutiae, get false minutiae or misidentify and so on. If the computer can automatically extract Leukocyte areas as ROI from cells image, medical examination staffs will skip the process of catching ROIs, and directly and easily judge the Leukocyte areas. Therefore, the proposed method with using computer image processing technology to automatically extract Leukocyte areas as ROIs in the bone marrow cells image is valuable in the fields of medical image processing.We will extract the Leukocyte areas based on Improved ITTI and Particle Swarm Optimization(PSO). In which, firstly, after using Gaussian filtering and Multi-scale normalization methods, the color salient features figure, brightness notable feature figure and the direction of the salient features figure are gained. And then the three feature figures are fused into a saliency figure. Finally, based on improved Particle Swarm Optimizing Otsu threshold, ROIs are extracted availably. This paper proposes mainly two improved methods as follows:(1)Through using ITTI model of Gaussian filtering and Multi-scale normalization,extracting the direction, brightness and color saliency characteristic from the original grayscale image, respectively. And then, according to the principle that contributions of the human's visual features are inconsistent, the original ITTI model with linearly combining with three significant features will be improved. This paper presents a new salient features figure with adaptive coefficients.(2)After gaining the salient features figure with employing adaptive coefficients, the saliency map of a ROI is got by the Otsu based on improved PSO, for extracting the full Leukocyte areas.Experiments show that this method is validated to enable extract the ROI in the bone marrow cells image. This algorithm improves the accuracy of extracting the ROI.