Yitao Liang

Preprocessing study of retinal image based on component extraction

Extracting the clear retinal image information is a prerequisite for diagnosing the patientpsilas conditions exactly in clinic. In order to extract the clear retinal image with less data as assistant diagnosis, the histogram equalization preprocessing of colored retinal image is researched based on its features. The clarity is observed and analyzed by comparing equalization images of the colored retinal imagepsilas gray image, red component image, green component image and blue component image. Equalization images of the green component image is the most clear one, followed by equalization images of the colored retinal image, red component image and blue component image. The green component image not only holds the basic feature of the original color image with its data compressed down to 1/3, but also is more convenient to acquire.

Research on the test target of MTF for the remote sensing camera

Its difficult to measure the MTF of the remote sensing camera accurately because of the registration error between the image of the target and the pixel of the CCD at the initial sampling point when using the rectangular as the target. To reduce the effect of the error, the trapezoidal target, which combines the rough division and the subdivision, is put forward. Then, the formula of the error MTFε which results from the registration error is deduced under the Nyquist frequency. From which, the influence of the precision of the rough division and the slope of the target on the measurement accuracy of the MTF is discussed respectively. Analytical results show that the influence of the registration error can be reduced effectively and the measurement accuracy of the MTF can be improved greatly when using the trapezoidal target.

Study of estimating method for actual size of underwater moving objects

A novel method to estimate actual size of underwater moving objects on condition of no any references was proposed. Defocusing effect resulted from different distance between moving objects and lens was analyzed. Blurred edge width in out-of-focus image was detected by using wavelet transformation. The relations between blurred edge width and real distance between objects and lens were deduced based on optical principles. The actual size of the moving object was derived from Gauss formula and lateral magnification expression. The results show that the method gives the actual size of objects and avoids visual error. The method had been applied in the certain international tournament and its effect was excellent.

The detecting technology for the transient feeble optical detection system

Studied in accordance with the technology of photoelectrical detection and detecting transient feeble optical signal, the thesis presents the application system for detecting. On analyzing signal properties, the detecting scheme based on FPGA technology is proposed with the systems core modules designed and their core parts selected. Relational software and hardware is fulfilled and debugged successfully. The test results with self-made simulative signal confirmed the feasibility of the design scheme and the reliability of the practical system.

Measurement of the image motion of the medical image based on the No-zero joint transform correlator

In clinical diagnosis and treatment, the images which we wanted are taken in obscuration or the object which is caught is moving during imaging, which will cause the SNR of the images to be very low, and make the image registration difficulty. To improve the quality of image quality, the joint transform correlator is used to measure the image motion between the serial images, then the images are aligned and added together to form a distinct one. The principle of this method was described firstly, to improve the accuracy of the measurement, the No-zero joint transform correlator is then put forward, and the simulation and experiment was done by using some medical images lastly. The results showed that, the joint transform correlator can measure the sub-pixel image shifts entirely even for the images with SNR=1, and mean square deviation of the measurement error no more than 0.2 pixels, which can meet the requirement of the registration of the image absolutely.

Application of the optical correlator in the medical image registration

Medical image registration, which is the precondition of medical image fusion, is a highlight of present medical image processing, and has important values in clinical diagnosis and treatment. To match the medical images with low SNR, the joint transform correlator is used to measure the image motion between the serial images. The principle of this method is described, then the experiment is done by using some medical images. The results showed that, the joint transform correlator can measure the sub-pixel image shifts entirely even for the images with SNR=1, and mean square deviation of the measurement error no more than 0.2 pixels.