| Image segmentation is a key problem in computer vision.In this paper we mainly studied the region-based image segmentation technique and cluster segmentation technique,and then carry out two new methods on segmentation.One method is that we can regard target as background using inverse segmentation according to the growth criteria providing by gray histogram.The experimental results showed that the accuracy of segmentation is up to 100%.By analysising it is showed that this method is suitable for the image segmentation, which the ratio of regard targe to the total segmentation area is relative small.The other mathod of segmentation is a fuzzy C-means image segmentation combined of edge detection to find a relatively independent area surrounded by closed edge using region growing method to complete the image segmentation according to the edge points Categories by physical principle of proximity.At the same time, we studied wavelet noise reduction algorithm in the image pre-processing.Wavelet transform can analysis the time - frequency at the same time.It can decompose the two-dimensional signal into different resolution scales, and thus it has been widely used in the area of the image de-noising.In this paper we mainly studied a method of wavelet threshold de-noising and carry out hyperbolic threshold function hyperbola .The results show that method proposed threshold of a new kind of hyperbolic function, experimental results show that the peak signal to noise in the new method is superior to the traditional method of denoising and soft and hard thresholding.Finally, a more comprehensive performance evaluation was also carried out on the new segmentation algorithm including the computational complexity analysis the new segmentation algorithm is also carried out a more comprehensive performance evaluation, including the computational complexity analysis, object-background area ratio, degree of human intervention, anti-noise and other evaluations. |
PDF Full Text Request