Abstarct: Several factors, such as imaging sensor impairments, electronic transmission of images, and compression operations, can lead to failures such as noise in the image. Noise are unwanted values that effect on the pixel values of the image. The noise on the image, in addition to reducing image quality, makes it difficult to process and analyze the image. Several methods have been proposed to eliminate image noise, but these methods mainly involve failures such as pixel value alignment and displacement. Our goal in this research is to improve the performance of image noise reduction method and reduce the damage caused by them. The first step to reduce noise effects on the image, identification on the type noise and its intensity. In this research, entropy and histogram images in low frequency areas are used to detect the type of noise. Using the image bitmaps, noise can be detected in the image. In each image, the intensity of noise in the image can be estimated using information such as mean and bitwise entropy variance. Attention to the human vision system is another important factor that can be effective in the performance of noise abatement methods. The human visual system has a higher sensitivity to non-planar areas in the low frequency areas of the image. Meanwhile, noise cancellation methods are also subject to malfunctions. Accordingly, noise cancellation is done in low frequency areas of the image more intensively than in non-obtuse areas. The method presented in this paper, using entropy information, pixon and clustering, divides the image into two low frequency and high frequency sections. Then, for noise reduction in each of these two regions, suitable components are selected for the filter. The proposed method also reduces the failure of noise cancellation filters while reducing the noise effect in the image and preserving the details and edges of the image. To evaluate the proposed method, IVC and CSIQ databaxses are used. Due to the lack of salt and pepper noise in these databaxses, two distinct databaxses including these types of noise are provided. The accuracy of the proposed method for detecting the noise type is 97.92%, and for determining the noise intensity with the negation of 0.2, 82.5%. The qualitative and quantitative results show that the proposed method, in comparison with previous methods, has a better performance in image noise reduction. In addition, using segmentation baxsed methode on the human vision system, can significantly improve the performance of the image noise reduction filters.