The number of bits it takes to represent an image is important  in storage and transmission. Color quantization, which was an active research field in the last two decades, is one way to reduce the memory requirements for images display and transmission. Reducing the image resolution, or using different image sampling techniques are alternative procedures that are used for image compression.

Our goal is to present a new non-uniform irregular image sampling technique, based on a combination of color (spectral) and resampling (spatial) quantization. We present an algorithm that  extends existing sampling techniques for gray level images to  color and video sequences. The result of the algorithm is dense sampling along the edges and sparse uniform samples in homogeneous domains. The algorithm explores the relation between the spectral and spatial information to achieve efficient image compression at low bitrate. We compare the results to other irregular sampling schemes and the JPEG algorithm.