I am wondering what zeros in CDT represent in the original image values, and why the matrix in the upper triangle form is better for compression.
yang991178
The zeros means that the basis patterns corresponding to those coefficients contribute very little to the details of the image. The reason that there are only values on the upper left part of the matrix is because the high frequency parts are likely to be wiped out from quantization, and is not necessarily true for every image. This matrix will later be reordered for better compression.
sunnyygou
Before this class, I never thought about why PNG is usually preferred over JPEG. PNG compression is lossless. JPEG resolution depends on the quality we choose. Throwing back to project 1 when we learned about downsampling. A lower JPEG quality here is just like the downsampling method.
ioannis-vm
It would be interesting to know what proportion of the overall compression is attributed to quantization, versus the compression achieved by chroma subsampling.
I am wondering what zeros in CDT represent in the original image values, and why the matrix in the upper triangle form is better for compression.
The zeros means that the basis patterns corresponding to those coefficients contribute very little to the details of the image. The reason that there are only values on the upper left part of the matrix is because the high frequency parts are likely to be wiped out from quantization, and is not necessarily true for every image. This matrix will later be reordered for better compression.
Before this class, I never thought about why PNG is usually preferred over JPEG. PNG compression is lossless. JPEG resolution depends on the quality we choose. Throwing back to project 1 when we learned about downsampling. A lower JPEG quality here is just like the downsampling method.
It would be interesting to know what proportion of the overall compression is attributed to quantization, versus the compression achieved by chroma subsampling.