Lecture 21: Image Sensors (85)
colinsteidtmann

Conclusion:

Point sampled: Simplest method. When downsizing an image, it simply picks a single pixel value from the original image to represent that area in the smaller image. For noise reduction: Not very effective. Noise is random. Picking a single pixel value might capture a noisy pixel, not necessarily an average representation of the area.

Averaged Down: More effective for noise reduction. When downsizing, it calculates the average color value of a group of pixels from the original image. This group corresponds to the area represented by a single pixel in the smaller image. By averaging, the random noise gets somewhat smoothed out.

aidangarde

It makes sense that the averaged-down image appears more smooth, but it must lose detail by averaging out pixel-wide partners. Is there a quantifiable way to measure the noise of the picture and determine whether averaged down or point sampled is the preferable method?

cvankeuren

Are there specific cases where point sampling yields a better image than averaged down? Imaging a world where cost/speed doesn't matter of course, I'm curious if averaged down always performs better.

TiaJain

To answer cvankeuren's question, I believe that point sampling might yield a better image than averaging down in cases where preserving sharp edges and fine detail is more important than reducing noise (so more useful in line art/images w/ distinct, high-contrast boundaries) where avging could result in unwanted blurring.

Zzz212zzZ

Taking more pixels into consideration (just as what averaged down does) is fairly a good strategy because it tends to store more information, instead, the point sampled strategy just throws these pixels away. By the way, the averaged pixel is more smooth.

You must be enrolled in the course to comment