www.heavensgatel2.fora.pl

[nlwxpearo]

Co-occurrence matrix method based on image texture classification


- Jiang Yue h. - ___ - N cockroach ∞. __ - ∞ d ---- Tomb of Bu Huan m light turned Pi ■. __ - Shoes din ∞ He 0__ - Saint - hl 'debate away from the 0 - Wall of Hope Pregnancy __--。 h . - __ - _ ∞ flap n. . . A __ - __ blanket ------ more friends who Tsui hg pregnancy 0H Rights ∞ b} limonene ---- cut the ball screw can gI site plan spines of Surveying and Mapping Mount 3 1996, the feature of array . Third, the fuzzy clustering analysis for image texture classification for image texture feature values for the purpose of the image texture classification method using fuzzy clustering analysis of fecal broadly divided image texture points to three steps: 1. Value of co-occurrence matrix texture features a variety of texture features pre-calculation methods,[link widoczny dla zalogowanych], so there may be a texture value of the whole small, while the other is larger overall,[link widoczny dla zalogowanych], this time in order to make the amount of texture feature vector roughly equal role, it is necessary the kind of small features on the overall value multiplied by a constant, or the kind of features of the overall larger values of certain multiples shrink, making the relationship of magnitude of each component will not be very poor. If the situation is acceptable for data preprocessing is not 2. Classification of objects measured by calculating the similarity value of Q for a given set of a fuzzy relation matrix R [,[link widoczny dla zalogowanych], if it satisfies 0] a 1, [r I], then we call R = [] for the fuzzy similarity matrix, it satisfies reflexivity and symmetry sex. Lot of fuzzy similar matrix method calculation, we selected the absolute value of index, ie: r. exp a Σl +,; (2) where (, J a 1,[link widoczny dla zalogowanych],2. ...,.) 3. The fuzzy similarity matrix, cluster calculated,[link widoczny dla zalogowanych], the obtained fuzzy similar matrix R-. ] .... We use direct clustering method to cluster analysis of the R matrix. The practice of this method is: for a given level set (∈: 0,1]), if r ≥, i think that the first sub-sample texture texture with the first sub-sample is the same type of texture; if ri2 <, then that the first sub-sample with the first texture, texture sub-sample does not belong to the same type of texture. So go on, plus taking into account the transfer of, all the samples can be sub-sample classification. Fourth, testing and analysis is based on the above model, we have 486 on the computer language developed by TurboC corresponding program. First test images were calculated for each sub-sample of co-occurrence matrix texture features of value; and then test to calculate the like, through the overview of the overall situation and past experience to determine the appropriate value of A; compared with the size, can reach the value of the fuzzy membership Results under the power of the image texture classification used in this experiment is nine internationally recognized standards of the most difficult to distinguish between image texture. The size of each piece? 50pixel × 750pixel, in the test sub-sample sizes are taken 100pixe [× 100pixe [. In order to verify the classification of the same kind of mistake will not produce different kinds of textures into the texture of the case. Our approach is from the texture of each piece of the standard image map with 20 sub-samples, each sample is divided into a set of sub-l0, so that was a total of l8 group. Table l lists the 18 sets of image textures characteristic values, Table 2 is the texture feature values of these fuzzy similar matrix R of the data. As can be seen from Table 2, when we choose a 0.9, Ⅲ (= 1,3,5,7,9,11,13, l5, 17) values were more than values, which match the actual situation, In fact i-i + l group and the first group of images are from the same kind of texture images. Most (jg-i +1) is less than, only r |】 a 0.904, 5is a 0.924, rB three kinds of images and video can not correctly distinguish between the ninth species, the other can be properly classified. V. CONCLUSIONS Our analysis of the existing co-occurrence matrix method based on image texture classification based on fuzzy clustering analysis discussed for this method. test showed that: ① the method is feasible and can effectively improve the efficiency of texture classification. ② like this method any number of pairs, and thus a broader applicability than the t test method, more practical. ③ The method is by value compared with the value for clustering. If so the value changes, then we can get a series of cluster classification. If the fixed value, and easy to unify the classification criteria.