Controllable and Real-Time Reproducible Perlin Noise

Perlin noise is widely used to render natural phenomena or enrich the variety of motion in computer graphics; however, there is less attention on controlling Perlin noise. We present an approach to modify and control the value of Perlin noise function, which closely follows a user-specified pattern while preserving the original statistical properties of the noise. The problem is formulated as a multi-level optimization process, in which the optimization is performed from low frequency to high frequency bands. Our approach can easily achieve global and local control in designing texture patterns and reproduce same patterns without re-optimization.