Image Compression Based on Artificial Neural Networks and K-means

Using artificial neural network is an effective method in image compression. Artificial neural networks have the ability to approximate the color space of an image by a smaller space than from the original image. The approximation will be easy if the input image has many similarities in color at different locations. However, input image has many different types of colors, the approximation becomes difficult. This paper proposes the image compression method using a neural network method combined with k-means to minimize the loss of color information of the image in the compression process. First, the proposed method split image into diferent blocks. Then cluster these blocks using k-means. Finally, this paper builds an approximation space using the neural networks for all groups of blocks. Experimental results on real images show that the proposed method is better than the conventional method