Moment invariant automatic threshold image segmentation algorithm based on gradient adjustment
Source: Application of electronic technology: Luo Fei Luo poetic way road Zhang Yanling Wang has
Abstract: An improved moment invariant automatic threshold algorithm is proposed. The algorithm is targeted to the shortcomings of the image detail for the moment constant automatic threshold method, and the gradient adjustment based on the target edge pixel is added based on the torque unchanged automatic threshold, so that the segmentation effect takes into account the overall and details of the image. The algorithm does not have iterative or search, and the calculation speed can meet the requirements of real-time. The simulation results show that the algorithm can effectively segment the target image. Keywords: Image Segmentation Threshold Selection Momorism Change Gradient Adjustment Target Tracking Image Segmentation is a key issue in computer vision, which is an important image analysis technology. Its purpose is to extract the meaningful features of the image or the feature that needs to be applied. The basic principle followed by image segmentation is that the features or properties inside the region are the same or similar. These features or properties are different in different regions, and there is a difference [1]. It is usually summarized as a threshold method based on gray histograms and two categories based on regional growth methods based on the gray histogram. Among them, the threshold method is the most widely used segmentation technique in image segmentation due to its simple, small computational amount, more stable. In recent years, with the proposal of the new theories and methods of various disciplines, many split techniques that combine specific theories, methods and tools have also proposed, such as splitting technologies based on mathematical morphology, using the split technology of statistical model identification methods. Network segmentation technology, etc. [2]. The image threshold segmentation is a process that separates the target from the viewground background based on a certain threshold. In actual systems, there is no distinct grayscale between image objectives and backgrounds, and as visible light illumination angles, the brightness of the target and the brightness of the background are changed. Therefore, the correct choice of threshold is important, directly affecting the accuracy of the segmentation and the correctness of the image description analysis. The adaptive threshold selection is usually used. Here, the gradient adjustment is described below, which overcomes the flaws of the torque unchanged automatic threshold method to achieve better segmentation. 1 Momorial constant threshold division method introduction moment is a mathematical characteristic of a random variable. The torque method is the parameter point estimate calculation method introduced by Karl Pearson in 1894. Its basic idea is that the sample extracted from the overall, and the moment of the sample reflects the overall moment to a certain extent. Therefore, the estimate of the sample torque function can be used as an estimate of the corresponding overall torque function. The torque method is a highly efficient normal test method. The specific method is: the sample moment is an estimate of the corresponding total torque; the estimate of the corresponding overall torque is the estimate of the corresponding overall torque. The most common application of this approach is to estimate the overall mathematical expectation using the sample average. From the statistical point of view, the image can be seen as a sample in a two-dimensional random process (random field), which can be seen as a fuzzy vision of the ideal image, which reflects the overall characteristics. From the perspective of statistics, the division is the characteristic of the sample estimate. The distribution of the sample is estimated, and the division itself is also a description and estimation of the whole. It is a parameter estimate, and the method of parameter estimation can be used. Perform the segmentation of the target image. The torque constant threshold segmentation is to divide the rectification of the image for the segmentation of the image. The basic idea is that the moment of the image before and after the threshold is divided. [3]. The torque constant threshold method can be seen as an image transformation that converts the original blur image into an ideal image. The k-position MK of the two-dimensional image is defined as: where i is a gradation value, pi is a pixel ratio of grayscale I in the image. For image splitting, if the two-value segmentation is performed, only two gray levels of ZO and Z1 after segmentation, and ZO
