We conclude our material on sampling signals with some "real world" examples. This video and the next few in the playlist investigate sampling and quantizing of both audio/music signals and images. In this first video we examine downsampling of an audio signal. An original WAV music file is loaded in Matlab, the signal is downsampled by some factor, and then the downsampled signal is reconstructed via sinc interpolation back to the original signal sampling rate. The downsampling process uses an anti-aliasing filter to prevent aliasing from occurring. As the signal is downsampled by a larger factor, more of the high-frequency content of the original signal is removed and thus the reconstructed signal looks "smoother" since it has less high-frequency content.

In the previous video we examined the impact of downsampling an audio signal. In this video, we examine the impact of changing the number of quantization levels used to store the signal amplitudes. An original WAV music file is loaded in Matlab that uses 8-bit quantization. The signal is then re-quantized to both 6-bit and 4-bit quantization levels and plotted to visualize the impact of fewer quantization levels on the signal. While signals with fewer quantization levels require less storage space, it comes at the cost of introducing additional quantization error.

In this third video we examine downsampling of an image. An original image file is loaded in Matlab, the image is downsampled by some factor, and then the downsampled signal is reconstructed via sinc interpolation back to the original image/pixel. The downsampling and reconstruction process uses an anti-aliasing filter to prevent aliasing from occurring. As the signal is downsampled by a larger factor, more of the high-frequency content of the original image is removed and thus the reconstructed image looks "smoother" since it has less high-frequency content.

In the previous video we examined the impact of downsampling and reconstructing an image. In this video, we examine the impact of changing the number of quantization levels used to store the image pixel values.



An original gray scale image file is loaded in Matlab that uses 8-bit quantization for each pixel (corresponding to 2^8 = 256 colors). The image is then re-quantized to 7-bit, 6-bit, 5-bit, .....quantization levels to visualize the impact of using fewer quantization levels on the image. While images with fewer quantization levels require less storage space, it comes at the cost of introducing additional quantization error as compared to the original image.