Bilateral Structure Tensor Based On 3 D Image Texture Analysis And Application

With the rapid development of3D imaging technology, it is widely applied to thegeoscience, biological and medical fields, which resulting a large number ofthree-dimensional images. The need of three-dimensional image processing and analysispromotes the development of this important research direction in three-dimensionalimage texture. This disert mainly study three-dimensional seismic images,3D seismictexture attributes are divided into geometrical properties and physical properties:geometrical properties usually associated with the geometry of the seismic, and they arevery effective on depicting the outline of the geological body, layered structure, the faultsystem; physical properties including the properties of the kinematics and dynamics,mainly refers to the amplitude, waveform, frequency, speed and attenuation,which isused to research lithology changes. Because of the geometric properties can be betterdescripping the fault and other information, so better describe and identify the reservoir,the paper mainly studies the geometric properties of3D seismic image.The typical geometric attributes include attributes such as coherence cube,inclination, orientation and curvature; This paper first reviews the geometric propertiesof the developments at home and abroad, and makes systematic classification of variousgeometric properties, from concept of the geometric properties, the extraction methodsthe actual data applications, several practical problems for geometric attribute extractionconducted in-depth study, formats a set of actual data processing geometric propertyextraction methods. Specific activities include the following:1. Taking into account the amount of data of the three-dimensional seismic imagesare very large (often tens of or hundreds of G bytes), during the calculation of geometricproperties, we need to adopt some effective methods to reduce the algorithm timecomplexity. Calculation for earthquake geometric properties using three-dimensionalsliding window mechanism, before and after two operations, many of the data repeateoperated, if the previous operator can retain repeat that part of the results directly for thenext operations.then it can dramatically improve the computing speed, Based on this,we propose a fast algorithm based on the sliding window, which greatly improves the usefulness of geometric attribute extraction method.2. For the reason that the calculation of the gradient structure tensor often useGaussian filtering window to smooth the noise, which often smooth out minorearthquake information, such as fissures, fissures information are imply the oil and gasreservoirs. Based on this, we propose a calculation method based on the bilateral filtergradient structure tensor which can effectively save the minor earthquake information,enabling more sophisticated description of hydrocarbon reservoirs.The proposed algorithms are applied to the Dutch North Sea F3work area, fromthe perspective of the treatment effect of the actual work area, both computationalefficiency and minor earthquake detection results have been greatly improved than themethods of geometrical properties.