Abstract: To make variational level set method applicable to the application of tongue segmentation, this article introduced a kind of color tongue image fast segmentation method based on level sets. The early variational level set method is improved from the following aspects: the boundary feature weight function is improved to make this method adaptable to color tongue image segmentation, and a kind of variable time step method is introduced to increase the efficiency and accuracy of segmentation. As the comparison experiment results show, our method is about 1 time faster than the early method, and its accuracy is higher than the early one.
Abstract: One of the main purpose of the watermark preprocessing is to improve the robustness and security. For this reason,this paper presents an image encryption algorithm, which combines position scrambling and gray scrambling scrambled according to Arnold transform.Then all of the pixels of each sub-block are scambled by the algorithm based on Logistic chaotic map.Finally, all of the Pixels are redistributed and scrambled totally.Basing on image location scrambling,it takes advantage of multi-dimensional Arnold transformation and Logistic chaotic map, image gray scrambling is achieved. By histogram analysis,key sensitivity anslysis and correlation analysis of adjacent pixels of the results of the simulation, indicating that the scrambling effect of the algorithm is good,and the key space is large.
Abstract: The digital watermarking technology is the most active topic in image protection recently. In this paper, a digital watermarking algorithm based on Wavelet transform has been developed, which is implemented in MEX from MATLAB to explore a new area of research in digital media.
Abstract: APTCHA is an important means to improve network security and prevent hackers. Since only a simple transfigure processing and the same style. CAPTCHA are lack of security and have a high rate to be cracked. Hence it is important for network security to design a new method of transfigure having random and complex feature. The paper creates an algorithm using contraction mapping principle and fixed point, with a way of circumvolution and translation. The algorithm randomly select a fixed point in the character picture and make a transfigure using fixed point principle. The experimental results show that the CAPTCHA based on contraction mapping has a strong randomness and difficult to be recognized by computer programs, which can provide a strong security for Web applications.
Abstract: In this paper, a fast background subtraction algorithm using codebook model is presented to extract moving objects from surveillance videos. The time for stopped objects being absorbed into the background can be controlled to deal with different applications and have nothing to do with the complexity of the scene. We implement the algorithm on GPU using CUDA, and optimize the implementation using pinned memory and asynchronous execution techniques. Experimental results are provided to demonstrate the accuracy, effectiveness, and efficiency of the proposed algorithm.
Abstract: This paper proposes a algorithm for detecting manual blur on images, which is usually used to remove obvious traces when tamper images. The algorithm first blurs the test image and blocks the both test image and blurred image. Then extracts and compares the sharp edge points in contourlet domain of the two images, so as to detect the suspicious blurred blocks. Furthermore, differences between manual blur and defocus blur can be indicated by our proposed method, and we can find out whether the image has been manual blurred. We establish a rich set of experimental images, and test results show that the average accurate detection rate is high, and the tampered regions can be always located. Our next work is to improve the robustness of the algorithm.
Abstract: We present a new algorithm to locate targets by matching image frames taken from a moving platform. We have noticed that an image point is environment sensitive, but those energy changes of grouped points have their own statistical similarities in two image frames within limited time interval. This approach analyzes correspondence of energy points around every feature points between inter-frames in image sequence in order to decide those feature points. Successful results are given for a vide frames.
Abstract: Correspondence estimation is one of the fundamental challenges in computer vision lying in the core of many problems, from stereo and motion analysis to object recognition. The direct matching method would computer the fitness for every pixel in the resolution space and it would sensitive to illumination change and other change. For the image attained by distance transformation can weaken the bad influence under geometric distortion and edge change, and have the ability to resistant inversion. The PSO algorithm supports parallel search of multiple points that are changed along a smooth trajectory within the search space. The paper applied distance transformation and PSO to image matching and the experiment results showed that comparing with the traditional direct matching, the method of the paper is not sensitive to illumination change and drastically reduced the required computation time.
Abstract: A new method of multiple watermarks embedding simultaneously to identify image status based on once and twice discrete fractional Fourier transform (DFRT) is presented. Random series watermark, chirp watermark and annotation watermark are embedded simultaneously in fractional Fourier transform domain to identify digital tamper detection, scanning detection, duplicate detection. According to the extracting situation of three kinds of watermarks from an image and the twice DFRT amplitude spectrum, we can know whether the image is tampered, scanned or duplicated. It provides a convenient way to identify the status of the image and will be an all-sided and comprehensive anti-counterfeiting technology.
Abstract: The principle of Zernike moments and the method of sub-pixel edge detection based on Zernike moments were introduced in this paper. With the consideration of the limitation of the sub-pixel edge detection algorithm by Ghosal, such as the lower location precision of the edge and the extracted wider edge than that of the original image, an improved algorithm was proposed. On the one hand, a mask of size nine multiply nine was calculated and could be applied for the edge detection. On the other hand, a new criterion for edge detection was put forward. Additionally, a series of experiments were designed and implemented. The experiment results show that accuracy of the improved algorithm is higher than that obtained from using other size templates and Ghosal algorithm.