Digital Image Inpainting Using Cellular Neural Network

Digital Image inpainting methods provide a means for reconstruction of small damaged portions of an image. Image or video resources are often received in poor conditions, mostly with noise or defects making the resources difficult to read and understand. Some methods are presented that can be used for the reconstruction of damaged or partially known images. We propose an effective algorithm with CNN, that can be used to inpainting digital images or video frames with very high noise ratio. Noises inside the cell with different sizes are inpainted with different levels of surrounding information. So, the result showed that an almost blurred image or unrecognized cell can be recovered with visually good effect. The proposed method takes the possibility of direct implementation of an existing CNN chip into account, in a single step, by using 3x3 dimensional linear reaction templates. This same method can be further used for processing motion picture with high percentage of noise.