Image resizing - bicubic downsampling

asked 2015-08-05 06:28:01 -0600

Eilg gravatar image

Dear OpenCV users,

I am downsampling an image by a factor of 5. I am wondering what



exactly do in this case?

Different from upsampling, for downsampling one does not want to interpolate the new pixel value at a certain position but average over pixel values. E. g.: a 5x5 patch is blue and the center pixel is red.

Linear or cubic interpolation would yield red for the new large pixel, but this is certainly not the desired effect, since all other pixels are blue. I can't find any documentation on how cubic and linear downsampling really work in this case. How is the surrounding aggregated to produce the new pixel?

Thanks and best regards,


edit retag flag offensive close merge delete