Abstarct: Calculating reservoir characterization and forecasting the production situation in different parts of reservoir and well, is an important task for petroleum engineer. with better recognition of affective factors in production, more accurate evaluation for development of the field can made. As regards carbonate reservoir are generally heterogeneous, their descxription and evaluation requires the use of special method and techniques. Identification of different types of porosity within a reservoir rock is a functional parameter for reservoir characterization since various pore types play different roles in fluid transport and also, the pore spaces determine the fluid storage capacity of the reservoir. Image analysis is baxsed on computer image analysis. The information obtained from this method is very fast. The reason of using this method is correct evaluate pore spaces distribution with digital images, which is used to predict the reservoir quality and performance. In this research, three intelligent models have been introduced using neural networks and support vector machine systems to separation and classify types of pores in thin sections which include intraparticle porosity, interparticle porosity, vuggy porosity, moldic porosity, biomoldic porosity and fracture, respectively. For this aim, firstly, thirteen geometrical parameters of pores of each image contains area, area/box, aspect, axis(major), axis(minor), box(x/y), diameter(mean), feret (mean), IOD, aspect ratio, roundness, size(length) and size(with) were extracted by image analysis techniques. Of these features and their corresponding pore spaces, 682 data were trained with three three SH-NN (Shallow neural network) and SAE (stacked auto-encoder) and support vector machine methods. Then the networks were tested with 277 test data, that way the network inputs, were the geometric properties of pores and outputs were different types of existing porosity in images. At the end, pore size distribution were estimated by them. The result show that the intelligent techniques have succeeded in predicting porosity types and the precision of support vector machine, deep neural network and shallow neural network models are %96.02, %92.05 and 81.58% respectively. In the following, the best classify accuracy of any types of pore spaces, was determined. Thus, the highest classification accuracy was obtained in all three systems, was the same amount 97.96% and 83.3% for intraparticle porosity and fracture respectively. For interparticle, vuggy and biomoldic porosities by support vector machine models, the accuracy 94%, 88% and 100% respectively, also for moldic porosity by two support vector machine and deep neural network models, was obtained the same accuracy 83.3%. In the next step, for review amount of porosity in the reservoir and its relation with the existing porosity; well logs is displayed, including total porosity, neutron, density and gamma, as well as calcite volume, volume of dolomite and shale volume, along with gas volume log at the depths of studied sections, with using geolog software, individually and briefly in three tracks. According to them, the highest achieved total porosity is 26%.