K-mer frequency of any DNA sequence is calculated by counting occurrences of all possible substrings of length k. The k-mer frequency of genome or next generation sequencing data is an invaluable tool to gain insights about the DNA sequence and its grammar. For genomes, k-mer counts can be used for motif discovery, classification and alignment-free comparison of multiple genomes. For short reads, k-mer counts are used for quality check, diagnosis, error correction and assembly. The initial step k-mer counting requires storage of frequency tables which tend to get bigger by increasing length of k. In this study we propose a method for lossless compression of k-mer data which is expected to simplify and facilitate storage and analysis of k-mer data. In a raster image, such as PNG, each pixel has two components; coordinate and color. We implemented Chaos Game Representation (CGR) to map k-mers to coordinates and k-mer occurrence was mapped to RGB color via bit-level operations. CGR maps can be divided and labeled according to the corresponding substring, each substring is mapped onto a sub-square, creating a fractal-like structure. Basically, the whole set of frequencies of the k-mers found in each genomic sequence are displayed in the form of a single image in which each pixel is associated with a specific k-mer and its occurrence. As result, file size has been reduced by approximately 10 times compared to plain text and reduced 5 times compared to binary storage (Jellyfish). Storage of k-mer data as image will not only save storage space but also facilitate genomic analysis in a manner previously not implemented. Image related algorithms can be used to process, analyze and manipulate collection of images representing genomic or next-generation sequencing data k-mer signatures.