Image Noise Estimation by Using Adaptive Image Block Division

An image was firstly divided into regular rectangle blocks in conventional block division based noise estimation methods, and then the noise intensity was calculated using the homogeneous blocks. However, it always leads to an over-estimation of the noise intensity when dealing with highly textured images. To solve this problem, we proposed a new noise estimation method based on an adaptive image block division algorithm. Firstly, the noisy image was decomposed into adaptive image blocks by using the spatial neighborhood pixels, where the adaptive blocks can adhere to local image structures, such as image edges and textures; then the most homogeneous adaptive blocks were detected with a scatter-plot-based statistical method; finally, the image noise intensity was calculated with the selected homogeneous adaptive blocks. Experimental results demonstrate that the proposed method is more accurate and stable than existing methods when estimating noise from images with various complexities and different noise levels.