Abstarct: Blurriness is one of the common distortions in images. This distortion is caused by spilling a pixel information over its adjacent pixels. Different reasons lead to various types of blurriness in the images. The knowledge about the type of image blurriness is one of the important parameters which directly affects the performance of the de-blurring methods. In this thesis, a method has been proposed to classify the type of blurriness in the digital images. Indeed, the proposed method can classify the type and the intensity of the image blurriness in the spatial domain both globally and locally. This method is a faster and more accurate method in comparison with the other exiting methods. In the proposed method the correlation concept is used to classify the type and the intensity of the image blurriness. The correlation concept depicts the effectiveness and the relations between the image pixels. Also, the model and the amount of correlation of adjacent pixels is proportional to the type of the blurriness. In the motion blur the correlation of pixels is only in a specific direction. However, in the other types of blur, the correlation of pixels is the same in the all directions. In the Gaussian blur, the effect of adjacent pixels depends on their distance from the considered pixel. The difference between the defocus and the rectangular blurs is the positions of effective pixels which are dependent on their blur kernel shapes. The shape of blur kernels in the defocus and the rectangular blurs are disk and rectangle, respectively. By increasing the intensity of image blurriness, the influence and the correlation of the adjacent pixels increases. Therefore, in the proposed method first correlation maps are obtained for the input image. Then, histogram-baxsed features of these correlation maps are used to detect the type and intensity of the blurriness. To evaluate the proposed method, a databaxse of blurred and non-blurred images has been created. The accuracy of the proposed method for classifying the blur type in global and local blur and for detecting the blur level is 90.4%, 80.1% and 91.2% respectively. Also, two most recently methods in spatial and frequency domains were implemented to evaluate the performance of the proposed method. The experimental results show that the proposed method has a better performance than the most recently methods in the terms of accuracy and speed.