You are viewing the course site for a past offering of this course. The current offering may be found here.
Lecture 22: Image Processing (10)
knguyen0811

The differences I notice between the original image and this one is the "glow", for lack of a better word, along the railing of the stairs.

Caozongkai

I honestly cannot tell the difference between the compressed CbCr channels and the original one. I thought after the down sampling the color or the color around the edge of the stairs would be inaccurate or blurred, but it seems that this technic works great.

surelywang

@caozongkai Agreed! I flipped back and forth between the two images and failed to spot a difference. Definitely a powerful technique.

fywu85

I wonder if there are cases when jpeg compression would work less effectively? Based on personal experience, I found it tends not to work as well when compressing fonts and diagrams.

michaeltu1

I think a more obvious difference between the images can be seen on the red tiles in the lower right of the image. The dirt on top of the tiles and the cyan color along the wall next to them are different.

frankieeder

Although this yields impressive results for the final image, if you need to do color correction beforehand, the chroma subsampled information is essential. Trying to color correct an already chroma subsampled image will cause noticeable effects, whereas first correcting then compressing will avoid these errors.

sandykzhang

Since humans are more sensitive to green, is it helpful to also compress Cb more than Cr?

jessicajyeh

Maybe the difference would be slightly more apparent on an image that is predominantly only color, since this image has most of its detail coming from the contrast of color against a dark black/brown color.

You must be enrolled in the course to comment